PDA

View Full Version : Pedigree Collapse Calculator



Tolan
01-03-2015, 02:27 PM
I tried this software:
http://www.y-str.org/2014/12/pedigree-collapse-calculator.html

I get 1.44%

And you?

Kaido
01-03-2015, 05:06 PM
1.92% - Fourth Cousins

ZephyrousMandaru
01-03-2015, 05:30 PM
4.36% Third cousins, disgusting.

DMXX
01-03-2015, 05:32 PM
0.44%, which would make my parents sixth cousins. Somewhat surprised it wasn't 0% given the geographical separation between their areas of origin.

Sangarius
01-03-2015, 06:29 PM
1,5% - that would make my parents fourth cousins, which I think is quite exaggerated.

While I do like playing around with genome calculators of all sorts and am grateful for people developing them and offering for free, my experience with the calculators from y-str.org is that they are most often very inaccurate.

vettor
01-03-2015, 06:41 PM
zero percent for me

Tolan
01-03-2015, 07:19 PM
While I do like playing around with genome calculators of all sorts and am grateful for people developing them and offering for free, my experience with the calculators from y-str.org is that they are most often very inaccurate.

I think this calculateur takes into account also the IBS segments even if they are not IBD.
That is why it gives a high rate.
That said, if your parents are of the same origin, the indenticall IBS segment will most certainly more.
On the other hand, if the population of origin of your parents has a high genetic variability, it is likely that the rate is smaller.

Probably the number given by the calculator is good, but not the interpretation ..

Salkin
01-03-2015, 08:03 PM
1.49% - fourth cousins. Dunno about that. GEDmatch's "parents related" utility says they are not.

vettor
01-03-2015, 09:46 PM
zero percent for me

edit this...used wrong file before

it is 1.68% ..................4th cousins

and my son is 0.78% ...............5th cousins

Scarlet Ibis
01-04-2015, 01:07 AM
My parents' ancestries are from generally the complete opposite ends of the Eurasian landmass, so no need for me to bother. ;)

ilmari
01-04-2015, 10:55 AM
Commenting here now to check when I am back in town. But I already have tons of pedigree collapse on my tree, and that is my maternal side, so calling this method pedigree collapse as well is going to put a fly into the ointment of the definition of pedigree collapse in general.

Tolan
01-04-2015, 11:11 AM
My parents' ancestries are from generally the complete opposite ends of the Eurasian landmass, so no need for me to bother. ;)

It would be interesting to see if you are close to 0%. Otherwise, this calculator is really bad!;)

Tolan
01-04-2015, 11:27 AM
Commenting here now to check when I am back in town. But I already have tons of pedigree collapse on my tree, and that is my maternal side, so calling this method pedigree collapse as well is going to put a fly into the ointment of the definition of pedigree collapse in general.
Thanks,
I speak very little English, and I did not know what it meant "pedigree collapse".
So, In French, it 's "implexe"
Totally different from consanguinity
Consanguinity of a person is only a small part of his pedigree collapse..

BMG
01-05-2015, 04:15 PM
It shows my parents are fifth cousins .1.05% .
I only know that they arent related for last three generations . beyond that is unknown to me but still they could be related since both of the families come from nearby villages

BMG
01-05-2015, 04:24 PM
It would be interesting to see if you are close to 0%. Otherwise, this calculator is really bad!;)
I too would like to see that . That would be a good test of this calculator.

Ignis90
01-05-2015, 04:33 PM
0.87%, which would be 5th cousins. My parents are from the same region but villages ~60km apart.

Gedmatch shows they aren't related though.

AJL
01-05-2015, 06:29 PM
It says the parents of my late grandfather, who was Ashkenazi, were 4th cousins with 1.73% overlap. This is plausible but it's more likely that they were from a restricted gene pool many hundreds of years ago, in other words, something like 20th-25th cousins 10 or more different ways.

This is palpably different than fourth cousins -- Gedmatch shows no long blocks over 7 cM with the same utility. Any actual 3rd-to-5th cousins I have connected with through FTDNA and 23andme either show >14 cM, or no overlap above 7 cM. Felix's incrementalist approach to IBD assumes that you can keep identifying segments to an asymptotically low amount but that is not the nature of recombination.

It also says my parents are fourth cousins (1.45%), however, which is clearly impossible. Gedmatch shows nothing, and they come from completely different ethnic backgrounds.

It says too that my mother's parents are fifth cousins, which is also impossible. Gedmatch again shows nothing. Two of her fourth great-grandparents were first cousins, though, and were from the remote Teesdale area of northern England, so were likely 7th cousins 4 different ways or summat. (Sorry, that's my inbred Yorkshire talking there. :P)

So I have to agree with others that (1) IBS is being confused with IBD, as Felix has done often before, and (2) interpretations are being made well beyond what is supported by available evidence.

felix
01-06-2015, 04:04 PM
It says the parents of my late grandfather, who was Ashkenazi, were 4th cousins with 1.73% overlap. This is plausible but it's more likely that they were from a restricted gene pool many hundreds of years ago, in other words, something like 20th-25th cousins 10 or more different ways.

This is palpably different than fourth cousins -- Gedmatch shows no long blocks over 7 cM with the same utility. Any actual 3rd-to-5th cousins I have connected with through FTDNA and 23andme either show >14 cM, or no overlap above 7 cM. Felix's incrementalist approach to IBD assumes that you can keep identifying segments to an asymptotically low amount but that is not the nature of recombination.

It also says my parents are fourth cousins (1.45%), however, which is clearly impossible. Gedmatch shows nothing, and they come from completely different ethnic backgrounds.

It says too that my mother's parents are fifth cousins, which is also impossible. Gedmatch again shows nothing. Two of her fourth great-grandparents were first cousins, though, and were from the remote Teesdale area of northern England, so were likely 7th cousins 4 different ways or summat. (Sorry, that's my inbred Yorkshire talking there. :P)

So I have to agree with others that (1) IBS is being confused with IBD, as Felix has done often before, and (2) interpretations are being made well beyond what is supported by available evidence.

Just stumbled upon this forum post, so I thought of clarifying the confusion.

No, I don't confuse IBS with IBD. The problem with IBD is, it only goes back upto the common ancestor of parents and any of the siblings of that common ancestor will not pass phased parents kits. I explained this effect on my blog Phasing: Good or Bad? (http://www.fc.id.au/2014/12/phasing-good-or-bad.html). Hence, IBD is perfectly alright for exogamous communities but for highly endogamous society like mine, where every marriage is a cousin marriage, IBD will never work.

Also, pedigree collapse calculation is not based on IBD or IBS. It is actually based on RoH. Pedigree collapse or cousin marriages actually destroys IBD because the shorter segments that appear on both parents from the same common ancestor will appear longer in the child, thus it is no longer IBD.

A single segment from the common ancestor (great-grandparent) will atleast have 8 potential recombination for first cousin parents, passing through two different lineages before appearing as double stranded RoH in the child. Hence, there is no way to restrict 7 cM since it will be broken down quite quickly.

I tested the tool based on 11 kits I manage - which do have first, second, third, fifth and the predicted values matches the expected value. The tool will be accurate for +/- 1 generation in general. However, it is to note, if the tool says parents are 5th cousin, then it could also mean parents are double 6th cousins, or quadruple 7th cousins and so on - since RoH works that way.

Hope this helps.

Felix

AJL
01-06-2015, 04:47 PM
The precipitous dropoff in DNA sharing around the level of fourth cousin means we cannot derive any useful information about degrees of kinship based on sharing of small segments. Endogamy does not change the rate of recombination, it just smudges the boundaries of what is IBD vs. IBS.

What you have to do with endogamous populations is set a higher threshold before treating a segment as IBD. This is what 23andme's and FTDNA's "Ashkenazi algorithm" is for.

In any case, matches below 13.28 cM are typically "beyond fourth cousin," (http://www.isogg.org/wiki/Autosomal_DNA_statistics) which is an exceedingly broad category that could mean anything from "approximately fourth cousins once removed" to "similar ethnicity" to "not a chimpanze" to "a fellow ape" to "also a living thing." You can find even common segments with a cucumber if you set the threshold low enough.

As I have linked to before (http://www.isogg.org/wiki/Identical_by_descent), the probability of a segment being IBS or IBD is .5 with a segment somewhere between 8 and 9 cM. The testing companies still include smaller segments, but questions arise with any segments less than about 13 or 14 cM: one cannot conclude with any certainty that they are IBD without phasing. Obviously the testing companies do not want to block real but distant matches, but your likelihood of finding a common genetic ancestor with a single segment of 8 cM is quite low, and your likelihood with your idea of a common segment (your conception of 4th or 5th cousins) is probably under .001.

Your objections to phasing and your use of endogamous sample sets to make broad inferences about exogamous populations are, quite frankly, why more people do not take you seriously.
Perhaps if you studied the work of well-versed people who have scientific backgrounds, like Dr. Tim Janzen, you would not be quite so confused.

dp
01-06-2015, 07:29 PM
Just stumbled upon this forum post ...
Welcome to Anthrogenica. Thanks for letting so many of us join the Archaic_DNA group. aDNA is a hot topic around here. So many people are curious not only how they compare to present populations but to archaic ones too.
Yours,
David Powell
dp :-)

ViktorL1
01-07-2015, 03:47 AM
1.12%. 5th cousins. I'm outbred?

Táltos
01-07-2015, 04:30 AM
The precipitous dropoff in DNA sharing around the level of fourth cousin means we cannot derive any useful information about degrees of kinship based on sharing of small segments. Endogamy does not change the rate of recombination, it just smudges the boundaries of what is IBD vs. IBS.

What you have to do with endogamous populations is set a higher threshold before treating a segment as IBD. This is what 23andme's and FTDNA's "Ashkenazi algorithm" is for.

In any case, matches below 13.28 cM are typically "beyond fourth cousin," (http://www.isogg.org/wiki/Autosomal_DNA_statistics) which is an exceedingly broad category that could mean anything from "approximately fourth cousins once removed" to "similar ethnicity" to "not a chimpanze" to "a fellow ape" to "also a living thing." You can find even common segments with a cucumber if you set the threshold low enough.

As I have linked to before (http://www.isogg.org/wiki/Identical_by_descent), the probability of a segment being IBS or IBD is .5 with a segment somewhere between 8 and 9 cM. The testing companies still include smaller segments, but questions arise with any segments less than about 13 or 14 cM: one cannot conclude with any certainty that they are IBD without phasing. Obviously the testing companies do not want to block real but distant matches, but your likelihood of finding a common genetic ancestor with a single segment of 8 cM is quite low, and your likelihood with your idea of a common segment (your conception of 4th or 5th cousins) is probably under .001.

Your objections to phasing and your use of endogamous sample sets to make broad inferences about exogamous populations are, quite frankly, why more people do not take you seriously.
Perhaps if you studied the work of well-versed people who have scientific backgrounds, like Dr. Tim Janzen, you would not be quite so confused.

AJL,
Thanks for the ISOGG link. Which has me thinking are the majority of my Jewish matches false?
Quote from the ISOGG link.

Smaller matching segments below 11 cMs are also more likely to be false matches. Many segments between 5 cMs and 10 cMs are likely to be false positives. For genuine smaller shared segments in the range of 5 cMs to 10 cMs the common ancestor may be as many as 10 to 15 generations or more back in time. It is particularly problematical to use small segments under 5 cM for genealogical matching purposes because very few of these segments will be true IBD matches. Durand et al (2014) analysed phased data from 2,952 father-mother-child trios in the 23andMe dataset and identified a false positive rate of over 67% for 2–4 cM segments.[5] The error rate is likely to be much higher in unphased data.

At FTDNA because they are the easiest to see (without having to share), my smallest Jewish match is 7.7 cm. My largest single segment match is for 20.58 cm. From what I can tell on 23andme it seems that a lot of them are around 9 cm. My top anonymous Jewish match is 3rd to 6th cousin 0.29%, 2 shared segments. Not sure what that equates out to in cm. So after finally having paper trail for my direct paternal line, who is the most likely source of these matches, there has been no Jewish found in the records. Not for the direct paternal line or any of the wives that married in. This is going back to 1782, and maybe even 1724. So 6 generations back for sure, and maybe to 8. Does this seem right? Are the matches real, and this Jewish ancestry is even farther back,? So far back, that I won't find were the conversion took place?

I'm pretty sure FTDNA has not applied their Ashkenazi algorithm for me. With 23andme do they apply it if you have the badge "Looks like you have some Ashkenazi Ancestry"?

BTW, I really enjoyed your cucumber analogy. Apparently the name of my paternal line's village means cucumbers! :biggrin1:

felix
01-07-2015, 07:59 AM
The precipitous dropoff in DNA sharing around the level of fourth cousin means we cannot derive any useful information about degrees of kinship based on sharing of small segments. Endogamy does not change the rate of recombination, it just smudges the boundaries of what is IBD vs. IBS.

What you have to do with endogamous populations is set a higher threshold before treating a segment as IBD. This is what 23andme's and FTDNA's "Ashkenazi algorithm" is for.

In any case, matches below 13.28 cM are typically "beyond fourth cousin," (http://www.isogg.org/wiki/Autosomal_DNA_statistics) which is an exceedingly broad category that could mean anything from "approximately fourth cousins once removed" to "similar ethnicity" to "not a chimpanze" to "a fellow ape" to "also a living thing." You can find even common segments with a cucumber if you set the threshold low enough.

As I have linked to before (http://www.isogg.org/wiki/Identical_by_descent), the probability of a segment being IBS or IBD is .5 with a segment somewhere between 8 and 9 cM. The testing companies still include smaller segments, but questions arise with any segments less than about 13 or 14 cM: one cannot conclude with any certainty that they are IBD without phasing. Obviously the testing companies do not want to block real but distant matches, but your likelihood of finding a common genetic ancestor with a single segment of 8 cM is quite low, and your likelihood with your idea of a common segment (your conception of 4th or 5th cousins) is probably under .001.

Your objections to phasing and your use of endogamous sample sets to make broad inferences about exogamous populations are, quite frankly, why more people do not take you seriously.
Perhaps if you studied the work of well-versed people who have scientific backgrounds, like Dr. Tim Janzen, you would not be quite so confused.

I am not sure what you are referring to, but pedigree collapse itself is a result of endogamous marriage, and any longer segment a child matches with CA of parents of an endogamous marriage cannot be an IBD by definition because it had gone through recombination. Hence, any segment that passes through a pedigree collapse is not an IBD because you can't have RoH without recombination.

I am not against phasing, in fact I wrote parent-less phasing, and phasing tools myself.

"common segments with a cucumber if you set the threshold low enough" ?? That's really disgusting! 23andMe v4 somehow assumes all missing inbetween SNPs in a matching segment (that's less than 0.5 cM if you are not aware of) from 23andMe v3 (nearly 50% of SNPs/half of the segment assumed) through imputation, and yet you don't think it can occur randomly (because it is 23andMe?) and the matching segment from 23andMe is bogus? Do you know exactly how imputation works? As mentioned, every single 23andMe v3 and v4 are matched through imputation (if you are not aware of). In imputation, IBS segments with much less than 0.5 to 1 cM is taken from population (totally unrelated to you) and assumed to be appearing in the missing SNPs in v4 that forms tiny segments which is then matched with v3 to get a matching segment. How ironic! You want to match a phased IBD and yet, ~50% of missing SNPs (tiny IBS segment) in 23andMe v4 is coming from an unrelated person and matched in 23andMe v3 and yet you blame me as if I know nothing.

I assume you have no idea what you are taking about with relation to IBD. Sorry to stumble upon and I believe it was a mistake. I did not come to discuss about me or my scientific background or to know if I am confused or not. I just came here to clarify a tool that I built in hope of being useful to others. Nothing else.

Tolan
01-07-2015, 11:12 AM
My results: 1,44% Fourth cousins.
If I take literally, that's wrong!
The closest relationship between my parents is 7th cousins (0,195%).

On the other hand, I can deduce something else.
For example, a child of first cousins:his father have got 50% grandparents identical with his mother. Or 50% of great-grandparents identical ect ..

To me, with 1.44% is 5.76% to the 5th generation of my parents or 5.76 to 6th ect ...
And 5.76% in 16th generation, at 500 years ago.
If my father has 5.76% of ancestors identical at the 16th generation with my mother does not surprise me (about 65,536 theoretical ancestors).

Am I right? I may not be very clear ...

felix
01-07-2015, 11:24 AM
For example, a child of first cousins:his father have got 50% grandparents identical with his mother. Or 50% of great-grandparents identical ect ..

For a first cousin parents, the common ancestor is great-grandparents and they pass only 25% of half identical to the child. Hence, only 12.5% of RoH.

jdean
01-07-2015, 02:54 PM
My results: 1,44% Fourth cousins.
If I take literally, that's wrong!
The closest relationship between my parents is 7th cousins (0,195%).

On the other hand, I can deduce something else.
For example, a child of first cousins:his father have got 50% grandparents identical with his mother. Or 50% of great-grandparents identical ect ..

To me, with 1.44% is 5.76% to the 5th generation of my parents or 5.76 to 6th ect ...
And 5.76% in 16th generation, at 500 years ago.
If my father has 5.76% of ancestors identical at the 16th generation with my mother does not surprise me (about 65,536 theoretical ancestors).

Am I right? I may not be very clear ...I'm definitely not clear, but there again I rarely am when it comes to autosomal DNA : )

According to Felix's tool my parents are predicted 4th cousins but also their parents are 4th cousins, I'm guessing that's not as inconsistent as it initially sounds ?

My parents were born in adjacent villages near Pontypool in Wales as were my paternal grandfather and maternal grandmother. My maternal grandfather was born only a little further away whilst my paternal grandmother was born in a town (also in Wales) about 20 miles away, so all said and done this result shouldn't be much of a surprise.

However in fact almost all my paternal ancestors going back 6 generations came from England and the odd couple that were Welsh lived about 50 miles to the north. These English ancestors were spread out over 5 counties along the SW and for the most part would have had limited ability to interact prior to meeting up in Wales.

My Mother's side conversely mostly came from this area or near (Monmouthshire) but I haven’t found any interconnection in that time frame and there are only a couple of lines that I'm stuck at 5 generations.

AJL
01-07-2015, 04:27 PM
At FTDNA because they are the easiest to see (without having to share), my smallest Jewish match is 7.7 cm. My largest single segment match is for 20.58 cm. From what I can tell on 23andme it seems that a lot of them are around 9 cm.

The large segment seems real enough by size alone, though one should also be aware that there could potentially be a few circumstances that could cause longer blocks to appear -- perhaps chromosomal inversions? Ann Turner might know. In any case, blocks around 7-9 cM are not likely to be traceable if the common ancestor involved, or the match, is Ashkenazi.

There is no visible flag I'm aware of that is associated with the Ashkenazi algorithms, but this brings up an important issue, which is one of reciprocality. In cases where a non-Ashkenazi person matches an Ashkenazi person, the companies need to decide whether a match ought to be treated to the algorithm or not, independent of whether a kit is.

From what I've seen, I suspect matches involving people with Ashkenazi ancestry should simply be subjected to a higher centimorgan cutoff, because if you treat all matches as Ashkenazi and prioirtize multiple segments, as both companies do, you will end up prioritizing endogamous matches, which in 23andme's case is a particular problem because of the limit to how many matches you have, and because Relative Finder populates in such a way that Ashkenazi or other endogamous matches will always be prioritized.

FTDNA is much better in this regard than 23andme but both could learn something from analyzing their own data sets and comparing this to known ancestries/genealogies.

I guess there's no definitive answer to your question but since you have Lithuanian ancestry, and Lithuanians were pagans much later than pretty well any other Europeans, I am tempted to guess you have parallel geographic (for Lithuanians) and cultural (for Ashkenazim) endogamy and the common ancestor could be either Lithuanian or Ashkenazi, or perhaps even of a third background very similar to one or the other of these groups. How is the segment painted with the various admixture tools?

AJL
01-07-2015, 04:43 PM
any longer segment a child matches with CA of parents of an endogamous marriage cannot be an IBD by definition because it had gone through recombination.

What do you mean by "CA"?

Anyway, defining things in a way that suits your beliefs does not change fact. An IBD is of course inherited from parents, and all DNA a child has is inherited from his or her parents, barring a few de novo mutations which represent a minuscule proportion of the genome. Therefore, segments from a parent are IBD.

After that, it gets slightly more complex but it's still patently clear that shorter segments are more likely to be IBD with fewer transmission events, e.g. a grandparent-grandchild pair with a 3.0 cM block is much more likely to be IBD than two complete strangers, but could also be IBS. This is what recombination does, and this is what I believe you don't understand and which has caused you to arrive at an erroneous conclusion.


Do you know exactly how imputation works?

Yes: it's a fancy word for "guessing." 23andme is reducing the number of SNPs on their latest chip and being forced to compare across 4 platforms with unequal coverage, at a time when other companies are increasing the density of their scans. Their phasing for Ancestry Composition appears to work fairly well for the most part but the routine existence of sub-0.5% runs of DNA that are ascribed to backgrounds not found by any other test, and often not matching genealogies, still illustrates that the short blocks they use may sometimes stll be unreliable. Still, Countries of Ancestry does appear to be mainly accurate.

{EDIT}
I'd also refer you to other work on imputation. For example, in cattle (http://www.ncbi.nlm.nih.gov/pubmed/24906026) the accuracy rate was ~.95 and "Imputation from lower- to higher-density genotype platforms...was generally greater when the reference population was solely comprised of the breed to be imputed compared to when the reference population comprised of multiple breeds." In essence, this means it is quite likely that 23andme's imputation is skewed by relatively high representation of customers with Colonial American and Ashkenazi ancestry.

Táltos
01-07-2015, 07:57 PM
The large segment seems real enough by size alone, though one should also be aware that there could potentially be a few circumstances that could cause longer blocks to appear -- perhaps chromosomal inversions? Ann Turner might know. In any case, blocks around 7-9 cM are not likely to be traceable if the common ancestor involved, or the match, is Ashkenazi.

There is no visible flag I'm aware of that is associated with the Ashkenazi algorithms, but this brings up an important issue, which is one of reciprocality. In cases where a non-Ashkenazi person matches an Ashkenazi person, the companies need to decide whether a match ought to be treated to the algorithm or not, independent of whether a kit is.

From what I've seen, I suspect matches involving people with Ashkenazi ancestry should simply be subjected to a higher centimorgan cutoff, because if you treat all matches as Ashkenazi and prioirtize multiple segments, as both companies do, you will end up prioritizing endogamous matches, which in 23andme's case is a particular problem because of the limit to how many matches you have, and because Relative Finder populates in such a way that Ashkenazi or other endogamous matches will always be prioritized.

FTDNA is much better in this regard than 23andme but both could learn something from analyzing their own data sets and comparing this to known ancestries/genealogies.

I guess there's no definitive answer to your question but since you have Lithuanian ancestry, and Lithuanians were pagans much later than pretty well any other Europeans, I am tempted to guess you have parallel geographic (for Lithuanians) and cultural (for Ashkenazim) endogamy and the common ancestor could be either Lithuanian or Ashkenazi, or perhaps even of a third background very similar to one or the other of these groups. How is the segment painted with the various admixture tools?
AJL,
Thank you for the detailed answer! Good idea about looking at the chromosome in question individually. I just gave someone this very advice in another thread.

When I look at MDLP-K23b here is what I get by breakdown%:
Caucasian-27.4%, EHG-21.7, EEF-16.7%, North African-10.2%, South Central Asian-9.2%, Near East-8.6%, East Siberian-3.6%, Melano_Polynesian-1.5%, others are below 0. something so I won't bother to put them.
This one is very hard to look at with chromosome painting as there are a lot of different populations and the colors some can be the same. I picked this one because in the three ancestor oracle it is really good to describe my known basic heritage of my four grandparents.

I'm pressed for time now, later I will put the World 9 Dodecad. But just by eyeballing the miniature chromosome painting, this one is easier to look at. There is large swath of Atlantic-Baltic, smaller amount of Southern, and then even smaller amounts of Caucasus-Gedrosia, around the area that my Jewish matches are.

BTW, I think the pagan would suit me just fine! :biggrin1:

dp
01-07-2015, 08:31 PM
AJL,
Thanks for the ISOGG link. Which has me thinking are the majority of my Jewish matches false?

In the case of Ashkenazi auDNA I suspect people pick up matches not in recent genealogical time.
At least one of my FamilyFinder matches lives in England. With a longest segment of 24cM he was estimated as 2nd Cousin - 4th Cousin, predicted 3rd cousin (my 44th top match). This was recently confirmed in near family of both parties testing at 23andme. His sister (I think) came up as my aunt's 348th match --with at least a 15cM segment, so in the top 3rd of matches as a 3rd to 6th Cousin (predicted 4th). All four grandparents were specified from England. My aunt also came up having over a 15cM segment with someone with all four grandparents in Germany too.
Has anyone scanned the emails of matches from GEDmatch. Besides .uk and .de, I see .au, .nz regularly. And oftentimes I come across .sw, .fi; with even .ru, and .no? Of course some of those segments could be IBS, due to the 7cM default.
dp :-)
PS: I guess I should have prefaced the above my stating that my ancestors so far have been within a few hundred miles of tidewater VA since the early part of the 1700s.

Wolfie
01-07-2015, 09:53 PM
Many Finns among my distant matches. Norwegians, Swedes, and Danes, too.

DMXX
01-07-2015, 10:08 PM
{EDIT}
I'd also refer you to other work on imputation. For example, in cattle (http://www.ncbi.nlm.nih.gov/pubmed/24906026) the accuracy rate was ~.95 and "Imputation from lower- to higher-density genotype platforms...was generally greater when the reference population was solely comprised of the breed to be imputed compared to when the reference population comprised of multiple breeds." In essence, this means it is quite likely that 23andme's imputation is skewed by relatively high representation of customers with Colonial American and Ashkenazi ancestry.

Very much so. This reasoning is fully in line with the academic reality. Imputation can be confidently completed for those with European heritage to begin with given the extensive amount of samples allowing us to make those a priori assumptions of allelic status. This is simply not the case with other world groups right now, particularly in Africa. That was (until very recently) part of the reason why medical genetics involving non-European populations didn't go very far.

Táltos
01-07-2015, 11:32 PM
World 9 admixture percentages for chromosome 7:
Atlantic_Baltic- 48%,Southern -27.3%,Caucasus_Gedrosia- 22.2%, South Asian-1.2%, Siberian and Australasian are both 0.6%

23andme-chromosome 7 is the one with the matches (this painting shows it the best)
3343

Miniature Dodecad-World 9
3344

MDLP-K23b
3345

I think in the MDLP one there is a lot of interesting colors in that area. :)

AJL
01-07-2015, 11:41 PM
World 9 admixture percentages for chromosome 7:

Thanks, it does look like there's a potentially Levantine block there. Still, a hypothetical conversion to Catholicism might date from the 16th century.

Apparently one frequent reason for conversion was to escape the rabbinical courts, which were granted extensive powers in Lithuania and Poland for much of their history -- but their jurisdiction was of course limited to Jewish people.

felix
01-07-2015, 11:52 PM
What do you mean by "CA"?

Anyway, defining things in a way that suits your beliefs does not change fact. An IBD is of course inherited from parents, and all DNA a child has is inherited from his or her parents, barring a few de novo mutations which represent a minuscule proportion of the genome. Therefore, segments from a parent are IBD.

After that, it gets slightly more complex but it's still patently clear that shorter segments are more likely to be IBD with fewer transmission events, e.g. a grandparent-grandchild pair with a 3.0 cM block is much more likely to be IBD than two complete strangers, but could also be IBS. This is what recombination does, and this is what I believe you don't understand and which has caused you to arrive at an erroneous conclusion.

CA is common ancestor.

It depends on who you want to call as complete stranger. If you restrict your pedigree for just 4 generations while your parents are first cousin, then your third cousin becomes a stranger since he does not share an IBD with you but an IBS and the matching segment will be recombined.

The fallacy of Identity by Descent (IBD) (http://www.fc.id.au/2015/01/the-fallacy-of-identity-by-descent-ibd.html)


Yes: it's a fancy word for "guessing." 23andme is reducing the number of SNPs on their latest chip and being forced to compare across 4 platforms with unequal coverage, at a time when other companies are increasing the density of their scans. Their phasing for Ancestry Composition appears to work fairly well for the most part but the routine existence of sub-0.5% runs of DNA that are ascribed to backgrounds not found by any other test, and often not matching genealogies, still illustrates that the short blocks they use may sometimes stll be unreliable. Still, Countries of Ancestry does appear to be mainly accurate.

{EDIT}
I'd also refer you to other work on imputation. For example, in cattle (http://www.ncbi.nlm.nih.gov/pubmed/24906026) the accuracy rate was ~.95 and "Imputation from lower- to higher-density genotype platforms...was generally greater when the reference population was solely comprised of the breed to be imputed compared to when the reference population comprised of multiple breeds." In essence, this means it is quite likely that 23andme's imputation is skewed by relatively high representation of customers with Colonial American and Ashkenazi ancestry.

Why then do you think those missing small IBS segments are taken from a 'complete stranger' and imputed within an IBD matching segment in 23andme v4 to match against a v3 ?

felix
01-08-2015, 12:06 AM
I'm definitely not clear, but there again I rarely am when it comes to autosomal DNA : )

According to Felix's tool my parents are predicted 4th cousins but also their parents are 4th cousins, I'm guessing that's not as inconsistent as it initially sounds ?

My parents were born in adjacent villages near Pontypool in Wales as were my paternal grandfather and maternal grandmother. My maternal grandfather was born only a little further away whilst my paternal grandmother was born in a town (also in Wales) about 20 miles away, so all said and done this result shouldn't be much of a surprise.

However in fact almost all my paternal ancestors going back 6 generations came from England and the odd couple that were Welsh lived about 50 miles to the north. These English ancestors were spread out over 5 counties along the SW and for the most part would have had limited ability to interact prior to meeting up in Wales.

My Mother's side conversely mostly came from this area or near (Monmouthshire) but I haven’t found any interconnection in that time frame and there are only a couple of lines that I'm stuck at 5 generations.

Keep watch for duplicated ancestors. Also, a 12.5% ROH representing first cousin parents could also mean double second cousins or quadruple third cousins.. and so on. The value of RoH is correct but the best interpretation I can come with is what the message box says.

AJL
01-08-2015, 12:23 AM
If you restrict your pedigree for just 4 generations while your parents are first cousin, then your third cousin becomes a stranger since he does not share an IBD with you but an IBS

Endogamy does not magically transform IBD segments into IBS. Segments that derive from a common ancestor, regardless how inbred, are IBD. The boundaries of such segments may be "fuzzier" (Ann Turner coined this term which I think is very apt) than for other segments, but you cannot arbitraily decide to redefine all segments that derive from shared ancestry as IBS when they are not.

What I suspect may be tripping you up is that a segment of a short length, when examining two people who are not known relatives, is generally assumed to be IBS. However, within the four generations, as you describe, it is absolutely routine for short segments under the usual threshold to be IBD.

The very use of the centimorgan scale in genealogy lies in the fact that the probability of a segment length sub ~14 cM being IBS is not static but rather increases with the number of transmission events. With many segments we cannot state categorically whether they are IBS or IBD because no intermediate persons exist for comparison. However when such a person or persons do exist, we can often prove whether that segment is sub-threshold, for the purposes of being taken seriously for genealogy, or we can at least substantiate that the segment, if real, is much shorter than thought. Crossovers do occur where the parent will have a much longer match than the child but as you increase the number of transmission events this becomes more uncommon.

I have nothing against you personally, and while you clearly know about mathematics and programming and have been working hard to contribute to the field, you simply don't appear to have an adequate understanding of the transmission of DNA, with the result that your calculators cannot always be taken seriously.

vettor
01-08-2015, 12:43 AM
Why then do you think those missing small IBS segments are taken from a 'complete stranger' and imputed within an IBD matching segment in 23andme v4 to match against a v3 ?

so, is it an issue when comparing 23andme v4 with v3 ? .....or with other companies

I tested 23andme v3 and son v4 ...we have 3552.3 matching with longest Cm of 214.5
with my ftdna test and his 23andme v4 test........we have 3587.1 with longest cm of 281.5

clearly this odd mixed testing companies must make a difference with the segments!

felix
01-08-2015, 12:54 AM
Endogamy does not magically transform IBD segments into IBS. Segments that derive from a common ancestor, regardless how inbred, are IBD. The boundaries of such segments may be "fuzzier" (Ann Turner coined this term which I think is very apt) than for other segments, but you cannot arbitraily decide to redefine all segments that derive from shared ancestry as IBS when they are not.

Yes it does and I believe that's exactly where I you miss the point.

Endogamy transforms IBD segments into IBS. During endogamy, IBD segment from a single common ancestor of cousin parents - goes separately and independently on a different lineage to each parent and then recombines in the child to form the same exact IBD segment as in the common ancestor, making it IBS on the child because the segment matches the common ancestor but not found in each parent separately.

The consequence of this is, the child's phased kits will match all descendants of the common ancestor of the cousin parents (because they are IBD and no recombination had happened in each separate lineage of the parent), but none of the relatives who are descendants of the sibling to that common ancestor, because they become IBS (and will not match phased kits) as they had undergone recombination in the child's parents.

AJL
01-08-2015, 01:17 AM
Yes it does and I believe that's exactly where I you miss the point.

Endogamy transforms IBD segments into IBS. During endogamy, IBD segment from a single common ancestor of cousin parents - goes separately and independently on a different lineage to each parent and then recombines in the child to form the same exact IBD segment as in the common ancestor, making it IBS on the child because the segment matches the common ancestor but not found in each parent separately.

The length of a segment may not be the same but the segment still comes from a parent, or both parents, and is therefore IBD. Put another way, a segment may be artificially elongated slightly when comparing a parent (X), a child (Y), and a mutual match (A), but we can be completely sure that all segments between X and Y are IBD because all DNA you have comes from your parents.

What I suspect you are getting at, if you are indeed getting at anything, is that segments of a given length appear in a child that don't appear in the same length in either parent. But that doesn't make them conclusively either IBS or IBD.

felix
01-08-2015, 01:48 AM
The length of a segment may not be the same but the segment still comes from a parent, or both parents, and is therefore IBD. Put another way, a segment may be artificially elongated slightly when comparing a parent (X), a child (Y), and a mutual match (A), but we can be completely sure that all segments between X and Y are IBD because all DNA you have comes from your parents.

What I suspect you are getting at, if you are indeed getting at anything, is that segments of a given length appear in a child that don't appear in the same length in either parent. But that doesn't make them conclusively either IBS or IBD.

Your explanation for IBD is not the definition mentioned for IBD, because, IBD requires the segment to be without recombination. Hence, if a matching segment comes from both parents, then it can't be IBD but IBS.

felix
01-08-2015, 01:45 PM
so, is it an issue when comparing 23andme v4 with v3 ? .....or with other companies

I tested 23andme v3 and son v4 ...we have 3552.3 matching with longest Cm of 214.5
with my ftdna test and his 23andme v4 test........we have 3587.1 with longest cm of 281.5

clearly this odd mixed testing companies must make a difference with the segments!

There are no issues with 23andMe v4 matches (for recent ancestry in their own database), since imputing has its own advantages (cost effective, good for business) and limitations (not very accurate for ancestry, requires higher thresholds for accurate matching and cannot work with lower thresholds, etc). But, there will be some issues in the accuracy GEDmatch matches (I need to appreciate GEDmatch guys to make v4 compatible but the downside is, the matches may not be as accurate as other kits for distant relatives) and FTDNA rejected their data in their database as not compatible (which truly isn't compatible).

As everyone knows, 23andMe v4 is not designed for ancestry but for heath related SNPs.

AJL
01-08-2015, 03:27 PM
IBD requires the segment to be without recombination. Hence, if a matching segment comes from both parents, then it can't be IBD but IBS.

I believe you're still missing the point. If the segment exists in a child and parent it's IBD.

While the length of an IBD segment may be shorter than it appears you cannot conclude that, simply because a segment has undergone recombination, it is completely IBS unless its entire length does not appear after phasing. If some smaller portion does persist it is only the portion that has undergone recombination that is IBS.

Perhaps you should read this page (http://www.isogg.org/wiki/Identical_by_descent).

Táltos
01-08-2015, 04:02 PM
There are no issues with 23andMe v4 matches (for recent ancestry in their own database), since imputing has its own advantages (cost effective, good for business) and limitations (not very accurate for ancestry, requires higher thresholds for accurate matching and cannot work with lower thresholds, etc). But, there will be some issues in the accuracy GEDmatch matches (I need to appreciate GEDmatch guys to make v4 compatible but the downside is, the matches may not be as accurate as other kits for distant relatives) and FTDNA rejected their data in their database as not compatible (which truly isn't compatible).

As everyone knows, 23andMe v4 is not designed for ancestry but for heath related SNPs.

Ann Turner on FTDNA had pointed out that the v4 chip is not as horrible as it's been implied by to be. I agree with her. http://forums.familytreedna.com/showthread.php?t=36611&page=2

All of my kits, except my Mom's are on v4. Yes there are some differences on Gedmatch between my FTDNA kit and 23andme one. But most are truly the same. Even some of the more distant ones that share ancestry with me in the same village in Italy hold up with both kits on Gedmatch. My new favorite calculator on Gedmatch-MDLP-K23b also gives me the very best reading in 3 ancestor oracle, using my v4 kit!

Táltos
01-08-2015, 04:20 PM
Thanks, it does look like there's a potentially Levantine block there. Still, a hypothetical conversion to Catholicism might date from the 16th century.

Apparently one frequent reason for conversion was to escape the rabbinical courts, which were granted extensive powers in Lithuania and Poland for much of their history -- but their jurisdiction was of course limited to Jewish people.

AJL,
Thank you for looking at it. Good to know then that conversion might have taken place in the 16th century. Larry Mayka had once told me he thought my matches might possibly go back to the Middle Ages/Renaissance period. He must be right! I will probably never be able to find the paper record for it. I am very grateful though that I now have a village for sure of were my father's family was from.

I have to wonder if how far back my matches are will be able to help determine anything with my brother's Y DNA test. Or vice versa? I was pretty bummed out when his SNP came back negative. One of his closest matches took the Big Y, and they are at a GD of 3 at 67. I know that his match did not know of any Jewish ancestry either last I heard in his paternal line. Off the top of my head if I'm remembering correctly, his paper trail goes back to 1765.

felix
01-08-2015, 04:40 PM
I believe you're still missing the point. If the segment exists in a child and parent it's IBD.

While the length of an IBD segment may be shorter than it appears you cannot conclude that, simply because a segment has undergone recombination, it is completely IBS unless its entire length does not appear after phasing. If some smaller portion does persist it is only the portion that has undergone recombination that is IBS.

Perhaps you should read this page (http://www.isogg.org/wiki/Identical_by_descent).

No, I don't miss the point and I believe you are messing up with the definition of IBD again.

Just because a segment exists in a child and a parent doesn't mean it is IBD. It's simply not true. It is explained that way for beginners to understand and goes only for 1 generation. An IBD segment is from a common ancestor between two matches passed along the lineage "without recombination". Hence, any segment that is recombined through cousin marriage will have longer segments matching nth cousins who are related through lineage prior to the pedigree collapse are labeled as IBS and will not match parents. And this segment is not from one parent but from both parents coming from a single common ancestor of them. This tool which calculates RoH which is an overlap of two IBD segments (in relation to common ancestor, parents and child) which becomes an elongated IBS segment (in relation to common ancestor, match and child).

People who simply don't understand how IBS segment matches in their DNA just ignore them as noise which is in-fact informing about a true relation prior to a pedigree collapse.

ArmandoR1b
01-08-2015, 05:01 PM
I used the Pedigree Collapse Calculator on a person whose paternal ancestors were from a completely different part of the world than the maternal ancestors and it showed them to be 5th cousins. This is entirely impossible.

vettor
01-08-2015, 05:11 PM
I used the Pedigree Collapse Calculator on a person whose paternal ancestors were from a completely different part of the world than the maternal ancestors and it showed them to be 5th cousins. This is entirely impossible.

It worked for me, a fourth cousin via my ( father) paternal great GM mothers father son to my ( mother ) maternal great great GF side.

Can I make it simpler?...........a union from circa 1770 produced some children, one child married into my maternal tree side and the other child married into my paternal tree side ...........I got this from BDM registers

ArmandoR1b
01-08-2015, 05:24 PM
It worked for me, a fourth cousin via my ( father) paternal great GM mothers father son to my ( mother ) maternal great great GF side.

Can I make it simpler?...........a union from circa 1770 produced some children, one child married into my maternal tree side and the other child married into my paternal tree side ...........I got this from BDM registers

If the program has an invalid result for multiple people then it is no good. I just tried it on 3 more people in the same situation as my first example. The parents ancestors are from different parts of the world. The parents for all of the kits showed to be related to each other within 5 generations when it is impossible.

AJL
01-08-2015, 06:18 PM
If the program has an invalid result for multiple people then it is no good. I just tried it on 3 more people in the same situation as my first example. The parents ancestors are from different parts of the world. The parents for all of the kits showed to be related to each other within 5 generations when it is impossible.

While any reasonable person comes to the same conclusion as you, there's obviously no use pointing this out to felix because he apparently prefers playing word games to understanding how DNA is transmitted.

{EDIT}
David Pike's excellent utility (http://www.math.mun.ca/~dapike/FF23utils/pair-comp.php) suffers from none of the problems that felix's tool does. You can simply compare two kits, and look for runs of homozygosity of any length.

felix
01-08-2015, 06:26 PM
Ann Turner on FTDNA had pointed out that the v4 chip is not as horrible as it's been implied by to be. I agree with her. http://forums.familytreedna.com/showthread.php?t=36611&page=2

All of my kits, except my Mom's are on v4. Yes there are some differences on Gedmatch between my FTDNA kit and 23andme one. But most are truly the same. Even some of the more distant ones that share ancestry with me in the same village in Italy hold up with both kits on Gedmatch. My new favorite calculator on Gedmatch-MDLP-K23b also gives me the very best reading in 3 ancestor oracle, using my v4 kit!

Thanks for letting me know. I just replied to her in the thread.

felix
01-08-2015, 06:35 PM
I used the Pedigree Collapse Calculator on a person whose paternal ancestors were from a completely different part of the world than the maternal ancestors and it showed them to be 5th cousins. This is entirely impossible.

I already mentioned the RoH % is correct but the interpretation of 5th cousin is the best I can get based on RoH. However, the RoH % could also mean different cousins levels arriving the same percentage. The relationship interpretation are not so accurate beyond 3rd cousins which is well known since RoH will have passed 8 potential recombination events and high probability of either loosing or gaining at each level.

felix
01-08-2015, 06:43 PM
While any reasonable person comes to the same conclusion as you, there's obviously no use pointing this out to felix because he apparently prefers playing word games to understanding how DNA is transmitted.

I just pointed out your inaccurate understanding of what an IBD is and you blame me for word games?
Do you still think just because a segment is found in a parent and a child is IBD? Don't you think it has to also occur in grandparent and great-grandparent up to the common ancestor?



{EDIT}
David Pike's excellent utility (http://www.math.mun.ca/~dapike/FF23utils/pair-comp.php) suffers from none of the problems that felix's tool does. You can simply compare two kits, and look for runs of homozygosity of any length.

Seriously? David Pike never interpreted his results.

vettor
01-08-2015, 06:54 PM
If the program has an invalid result for multiple people then it is no good. I just tried it on 3 more people in the same situation as my first example. The parents ancestors are from different parts of the world. The parents for all of the kits showed to be related to each other within 5 generations when it is impossible.

Are you concerned by the term "5th cousin" ?

I ignore all these cousins, be it 2nd, 3rd, 4th etc, because I found in most programs they mean not what is described

AJL
01-08-2015, 07:03 PM
Are you concerned by the term "5th cousin" ?

I ignore all these cousins, be it 2nd, 3rd, 4th etc, because I found in most programs they mean not what is described

No, what he's concerned with is that every person in the world, according to this tool, has parents who are 5th cousins or closer.

vettor
01-08-2015, 07:54 PM
an interesting article of a similar matter

http://dna-explained.com/2015/01/02/how-phasing-works-and-determining-ibd-versus-ibs-matches/

felix
01-08-2015, 10:50 PM
No, what he's concerned with is that every person in the world, according to this tool, has parents who are 5th cousins or closer.

This is one of the reasons, FTDNA (and others) simply interpret as '5th to remote' cousins. The point is, if a person has RoH then their parents are related within genealogical timeframe (< 500 years).

Given the fact even the royal family has a non-parental event not recorded in their pedigree (Richard III) and Diana's maternal line was also not mentioned correctly as an Indian slave, while maintained by professional genealogists for hundreds of years, it is certainly possible most pedigree had little secrets that hides what future generations don't know to know. That's in every pedigree. I suggest to trust the RoH in the DNA rather than known pedigree which is been mostly hearsay beyond living people. (1.5% of your DNA is approx ~28 Mb of total homozygous segements). 

It also doesn't matter if someone is on the other side of the world or 'unrelated' unless you check the DNA. Every part of the world is contacted during colonial rule. Do you think 1.5% of RoH (which is 28 Mb) somehow magically appeared in DNA for so called unrelated people ?

AJL
01-08-2015, 11:00 PM
This is one of the reasons, FTDNA (and others) simply interpret as '5th to remote' cousins. The point is, if a person has RoH then their parents are related within genealogical timeframe (< 500 years).

Given the fact even the royal family has a non-parental event not recorded in their pedigree (Richard III) and Diana's maternal line was also not mentioned correctly as an Indian slave, while maintained by professional genealogists for hundreds of years, it is certainly possible most pedigree had little secrets that hides what future generations don't know to know. That's in every pedigree. I suggest to trust the RoH in your DNA than known pedigree which is been mostly hearsay. (1.5% of your DNA is approx ~28 Mb of total homozygous segements). 

It also doesn't matter if someone is on the other side of the world or 'unrelated' unless you check the DNA. Every part of the world is contacted during colonial rule. Is 1.5% of RoH is 28 Mb somehow magically appeared in your DNA?

You are clearly not capable of rational thought if you think everyone in the world is a fifth cousin to everyone else, and so I refuse to engage in any further conversation with you.

vettor
01-09-2015, 12:42 AM
No, what he's concerned with is that every person in the world, according to this tool, has parents who are 5th cousins or closer.

you can only carry a certain amount of genes from each side and these cease at the Supposed "5th cousin".
You carry nothing after that

there is a chart on this percentages that one carrys from generation to generation.

AJL
01-09-2015, 01:02 AM
there is a chart on this percentages that one carrys from generation to generation.

Yes, which I previously posted. To assert the existence of a measurable level of relatedness called "fifth cousins" which is detectable with unphased data, and at which all humans appear to be equally related, is bogus. All humans are not equally related, nor are they equally endogamous, and there is no use in a tool that suggests that this is the case. If the tool had integrity it would say something like "your parents are not related at the level of fourth cousins or closer." That is decidedly not the same as saying "your parents are related at the level of fifth cousins."

felix
01-09-2015, 01:58 AM
You are clearly not capable of rational thought if you think everyone in the world is a fifth cousin to everyone else, and so I refuse to engage in any further conversation with you.

Did I say 'everyone in the world is a fifth cousin to everyone else'? Did the tool even say that? No. It simply calculates RoH in DNA and displays the percentage and guesses the relationship. I already mentioned this several times. If you disagree with ROH in present in the DNA, the tool cannot "fit" what you think should be correct.

"not capable of rational thought" - Thanks for the feedback.

One small request: When you are having a conversation or disagreement over a topic, never comment personally about others because it truly hurts.

vettor
01-09-2015, 02:14 AM
Yes, which I previously posted. To assert the existence of a measurable level of relatedness called "fifth cousins" which is detectable with unphased data, and at which all humans appear to be equally related, is bogus. All humans are not equally related, nor are they equally endogamous, and there is no use in a tool that suggests that this is the case. If the tool had integrity it would say something like "your parents are not related at the level of fourth cousins or closer." That is decidedly not the same as saying "your parents are related at the level of fifth cousins."

Let me clear this up, I have no opinion on the tool either way.
I base these "cousin numbers" as fallicy because there is a finite end to the genes ones gets from generation to generation. the term 3rd, 4th or 5th cousin is a fantasy name IMO. I want to know if this program will find any relatives in my paternal half of my tree with my maternal half of my tree.
For me , as stated before, I do have a "union" which is inside the range of finite genes area..........I care little on what its called, because as I said, no company who tests DNA does a proper job of naming "cousins".

BMG
01-09-2015, 02:41 AM
Felix, A tool is as good as its interpretation .Even if the methodology is right wrong inferences can get users confused and that will invite criticisms . Please do take the criticism positively .Most of us criticizing here can't make tools half as good as yours . But most of the replies are honest and we are not convinced . You could take this as a feedback and make an even better one.

ArmandoR1b
01-09-2015, 03:16 AM
I already mentioned the RoH % is correct but the interpretation of 5th cousin is the best I can get based on RoH. However, the RoH % could also mean different cousins levels arriving the same percentage. The relationship interpretation are not so accurate beyond 3rd cousins which is well known since RoH will have passed 8 potential recombination events and high probability of either loosing or gaining at each level.

Then the tool is only good for people that actually have parents as 3rd cousins. That should be explicitly stated on the page that has the tool. There are adoptees and children that didn't know their father that could be led astray. It is unethical not to announce the caveat in simple wording that everyone can understand.


This is one of the reasons, FTDNA (and others) simply interpret as '5th to remote' cousins. The point is, if a person has RoH then their parents are related within genealogical timeframe (< 500 years).

Given the fact even the royal family has a non-parental event not recorded in their pedigree (Richard III) and Diana's maternal line was also not mentioned correctly as an Indian slave, while maintained by professional genealogists for hundreds of years, it is certainly possible most pedigree had little secrets that hides what future generations don't know to know. That's in every pedigree. I suggest to trust the RoH in the DNA rather than known pedigree which is been mostly hearsay beyond living people. (1.5% of your DNA is approx ~28 Mb of total homozygous segements). 

It also doesn't matter if someone is on the other side of the world or 'unrelated' unless you check the DNA. Every part of the world is contacted during colonial rule. Do you think 1.5% of RoH (which is 28 Mb) somehow magically appeared in DNA for so called unrelated people ?

The chances that the 3 different people that showed the parents to be 5th cousins to have a common ancestor in the past 500 years is 0. Not all European colonies had people from all the world. People from some parts of Europe dd not go to any colony until very recently. You have tunnel vision with the English, Spanish, French, and Dutch colonies. Not all Europeans are from those host regions and there are a lot of European countries that did not have colonies. Not all Europeans are related in the past 500 years.



Felix, A tool is as good as its interpretation .Even if the methodology is right wrong inferences can get users confused and that will invite criticisms . Please do take the criticism positively .Most of us criticizing here can't make tools half as good as yours . But most of the replies are honest and we are not convinced . You could take this as a feedback and make an even better one.

I agree. Felix has done some great things for the genealogical community and should be highly commended for that work by all of us. However, some of the points he tries to make are beyond logic and need to be argued until the he understands his error.

vettor
01-09-2015, 05:15 AM
The chances that the 3 different people that showed the parents to be 5th cousins to have a common ancestor in the past 500 years is 0. Not all European colonies had people from all the world. People from some parts of Europe dd not go to any colony until very recently. You have tunnel vision with the English, Spanish, French, and Dutch colonies. Not all Europeans are from those host regions and there are a lot of European countries that did not have colonies. Not all Europeans are related in the past 500 years.


The people that migrated from year 1900 to present makes your statement very strange, since the Bulk of migration in this period is the far greater percentage of the period that covers each person genetic inheritance. I have a union with a person in canada which is half way around the world, this person left my paternal area over 60 years ago, yet the "union " was circa 1895 which if we take it as 31 years per generation ( my calculation ) would make it between the 4th and 5th cousin .

felix
01-09-2015, 08:26 AM
I parameterized the tool, so that the ROH threshold settings can be entered manually.

Link: http://www.y-str.org/2014/12/pedigree-collapse-calculator.html

felix
01-09-2015, 09:19 AM
The chances that the 3 different people that showed the parents to be 5th cousins to have a common ancestor in the past 500 years is 0. Not all European colonies had people from all the world. People from some parts of Europe dd not go to any colony until very recently. You have tunnel vision with the English, Spanish, French, and Dutch colonies. Not all Europeans are from those host regions and there are a lot of European countries that did not have colonies. Not all Europeans are related in the past 500 years.

Europe grew 20 times it's population in the past 500 years and it has one of the highest pedigree collapse, without which you cannot explain the population explosion. For a European who had never set foot outside to match an another non-European person, it is enough if any of the descendant of the common ancestor admixed with that population.

ArmandoR1b
01-09-2015, 01:07 PM
The people that migrated from year 1900 to present makes your statement very strange, since the Bulk of migration in this period is the far greater percentage of the period that covers each person genetic inheritance. I have a union with a person in canada which is half way around the world, this person left my paternal area over 60 years ago, yet the "union " was circa 1895 which if we take it as 31 years per generation ( my calculation ) would make it between the 4th and 5th cousin .

Could you explain what you are saying in different words?

ArmandoR1b
01-09-2015, 01:26 PM
Europe grew 20 times it's population in the past 500 years and it has one of the highest pedigree collapse, without which you cannot explain the population explosion. For a European who had never set foot outside to match an another non-European person, it is enough IF any of the descendant of the common ancestor admixed with that population.

A population explosion in the past 500 years does not mean that a person from Ireland and a person from Yugoslavia (I'm using a hypothetical for the countries of origin) have a common ancestor in the past 500 years. Maybe in the past 2,000 or 4,000 years but not in the past 500 years. Ireland and Yugoslavia did not have colonies. Yugoslavs didn't migrate to Ireland and is it is extremely doubtful that Irish emigrated to Yugoslavia. If they both have trees going back at least 5 generations without any hint of either person have ancestry from the other country it is extremely doubtful that they have a common ancestry in the past 500 years and less doubtful that it just so happens that there is enough homozygous DNA from a common ancestor from 500 years ago to cause the calculator to think it is from a 5th cousin marriage. That's just ridiculous.

That is one huge IF by the way.

parasar
01-09-2015, 03:35 PM
A population explosion in the past 500 years does not mean that a person from Ireland and a person from Yugoslavia (I'm using a hypothetical for the countries of origin) have a common ancestor in the past 500 years. Maybe in the past 2,000 or 4,000 years but not in the past 500 years. Ireland and Yugoslavia did not have colonies. Yugoslavs didn't migrate to Ireland and is it is extremely doubtful that Irish emigrated to Yugoslavia. If they both have trees going back at least 5 generations without any hint of either person have ancestry from the other country it is extremely doubtful that they have a common ancestry in the past 500 years and less doubtful that it just so happens that there is enough homozygous DNA from a common ancestor from 500 years ago to cause the calculator to think it is from a 5th cousin marriage. That's just ridiculous.

That is one huge IF by the way.

It would be interesting to see if Italians show that same type of collapse.

The Geography of Recent Genetic Ancestry across Europe
http://arxiv.org/pdf/1207.3815v5.pdf

individuals from opposite ends of Europe are still expected to share millions of common genealogical ancestors over the last 1000 years...Some of the lowest levels of common ancestry are seen in the Italian and Iberian peninsulas...

We find that even geographically distant Europeans share ubiquitous common ancestry even within the past 1000 years...

There is relatively little common ancestry shared between the Italian peninsula and other locations, and what there is seems to derive mostly from longer ago than 2500ya...

Patterns for the Iberian peninsula are similar, with both Spain and Portugal showing very few common ancestors with other populations over the last 2500 years. However, the rate of IBD sharing within the peninsula is much higher than within Italy ...

...conclude that pairs of individuals across Europe are reasonably likely to share common genetic ancestors within the last 1000 years, and are certain to share many within
the last 2500 years...

... by 1000 years ago everyone (who left descendants) would be an ancestor of every present day European...

The fact that most people alive today in Europe share nearly the same set of (European, and possibly world-wide) ancestors from only 1000 years ago...someone in Spain may be related to an ancestor in the Iberian peninsula through perhaps 1000 different routes back through the pedigree, but to an ancestor in the Baltic region by only 10 different routes, so that the probability that this Spanish individual inherited genetic material from the Iberian ancestor is roughly 100 times higher. This allows the amount of genetic material shared by pairs of extant individuals to vary even if the set of ancestors is constant.

significant old substructure and large population sizes within Italy, strong enough that different groups within Italy share as little recent common ancestry as other distinct, modern-day countries, substructure that was not homogenized during the migration period. These patterns could also reflect in part geographic isolation within Italy as well as a long history of settlement of Italy from diverse sources ...

Their IBD:

1.1 Definitions: Genetic ancestry and identity by descent ... Fig 1 ...
If a pair of individuals have both inherited some genomic region from a common ancestor, that ancestor is called a “genetic common ancestor”, and the genomic region is shared “identical by descent” (IBD) by the two. Here we define an “IBD block” to be a contiguous segment of genome inherited (on at least one chromosome) from a shared common ancestor without intervening recombination (see figure 1A). A more usual definition of IBD restricts to those segments inherited from some prespecified set of “founder” individuals (e.g. Fisher, 1954; Donnelly, 1983; Chapman and Thompson, 2002), but we allow ancestors to be arbitrarily far back in time. Under our definition, everyone is IBD everywhere, but mostly on very short, old segments (Powell et al., 2010).

AJL
01-09-2015, 04:13 PM
^ For Italy, regional variation in endogamy rates was well illustrated by Cavalli-Sforza:

http://www.chapters.indigo.ca/books/consanguinity-inbreeding-and-genetic-drift/9781400847273-item.html

dp
01-09-2015, 05:55 PM
This question came to my mind after reading one of Vetttor's.
My mother tested with FamilyTreeDNA. Her sister with 23andme. When I put them onto GEDmatch --on one-to-many, they came up as 2652.1 cM auDNA shared, 167.2 cM longest segment, generation difference of 1.2 gens. Shouldn't they share more auDNA? I ran one-to-one and the total of segments > 7cM came to 2644.5cM which seems consistent.
I've checked at ISOGG and I see that the "actual half-identical value is 3338.31cMs" for sharing between parents to children, and siblings.
Below it they give a 2640cM value for full siblings but my impression is that would be the amount reported by 23andme or FamilyTreedNA, if they were on the same company, "since they do not report fully identical regions". The ISOGG section reads "FIRs can also be seen if you upload your raw data to the free GEDmatch utility. FIRs show up as "green" in the one-to-one comparison, though in the one-to-many chromosome browser display at GEDmatch, it only shows the regions as HIRs."
Well, I'm using GEDmatch, in one-to-one comparison. I'll give them that about half of the segments indicated by GEDmatch as a match are in green, and the other half in yellow, BUT together they do not take up all the chromosome lengths for all of them. Shouldn't GEDmatch give their matching in the area of 3338cM ?
In other words, AT GEDMATCH HOW MUCH DO FULL SIBLINGS COME UP AS, AND HOW MUCH DO HALF-SIBLINGS???
ISOGG says the later should have 1700cM sharing. That makes since if we were dealing with .5*3338cM rounded. But since I'm getting something in the 2650cM sharing range at GEDmatch, it seems to me that they are too related to be half-siblings.
Befuddled by GEDmatch,
dp

Táltos
01-09-2015, 07:04 PM
Befuddled by GEDmatch,
dp

I love Gedmatch, I truly do. Lately though, I'm not sure what is going on there. I'm sorry if this is off topic, or maybe related? But I have noticed especially with my FTDNA kit a lot of people who showed matching in the past, now show nothing when I put them in chromosome browser. They are known matches at the commercial companies too, though not all. Some are in the 10 cm above range yet some that are below 10 cm still show, not sure why?

Anyway to the best that I can list this. My distant cousin and I share ancestry from the same small village in southern Italy. It's a small Albanian mountain village. We all match on chromosome 3. My cousin, her father, and my mother are all on v3. I'm the odd duck on v4.

Comparison:
Father to me-23andme-188000000-191000000 5.7 cm 786 SNPs
on Gedmatch to my v4-190099633-193616397 8 cm 807 SNPs
Gedmatch to myFTDNA-190104243-193552969 7.8 cm 888 SNPs

Father to my Mom-23andme-188000000-192000000 6.9 cm 858 SNPs
on Gedmatch - 190093852-193998204 8.8 cm 1417 SNPs

His daughter to me-23andme-188000000-191000000 5.9 cm 807 SNPs
on Gedmatch-v4 190099633-193616397 8 cm 807 SNPs
Gedmatch to FTDNA kit- 190104243-193618300 8 cm 918 SNPs

His daughter to my [email protected] 6 cm 813 SNPs
on Gedmatch 190093852-193639869 8.1 cm 1348 SNPs

So I just put all this to show distant matches and how the breakdown is. I still agree with Ann Turner that we are dealing with "small fuzzy boundaries" so not a real deal breaker.

Here is my v4 kit to my mother's on Gedmatch-Total cm-3547 largest cm-214.5
My FTDNA kit to her on Gedmatch- Total cm-3571.7 largest cm-209.3

On FTDNA- 3377.33 total cm largest block-267.21
On 23andme-3719 cM total largest block-284 cM

In general Gedmatch shows everything to be larger for the most part. Hopefully this is not so hard for everyone to look at.

dp
01-09-2015, 08:24 PM
I love Gedmatch, I truly do. Lately though, I'm not sure what is going on there. I'm sorry if this is off topic, or maybe related? But I have noticed especially with my FTDNA kit a lot of people who showed matching in the past, now show nothing when I put them in chromosome browser. They are known matches at the commercial companies too, though not all. Some are in the 10 cm above range yet some that are below 10 cm still show, not sure why?

Anyway to the best that I can list this. My distant cousin and I share ancestry from the same small village in southern Italy. It's a small Albanian mountain village. We all match on chromosome 3. My cousin, her father, and my mother are all on v3. I'm the odd duck on v4.

Comparison:
Father to me-23andme-188000000-191000000 5.7 cm 786 SNPs
on Gedmatch to my v4-190099633-193616397 8 cm 807 SNPs
Gedmatch to myFTDNA-190104243-193552969 7.8 cm 888 SNPs

Father to my Mom-23andme-188000000-192000000 6.9 cm 858 SNPs
on Gedmatch - 190093852-193998204 8.8 cm 1417 SNPs

His daughter to me-23andme-188000000-191000000 5.9 cm 807 SNPs
on Gedmatch-v4 190099633-193616397 8 cm 807 SNPs
Gedmatch to FTDNA kit- 190104243-193618300 8 cm 918 SNPs

His daughter to my [email protected] 6 cm 813 SNPs
on Gedmatch 190093852-193639869 8.1 cm 1348 SNPs

So I just put all this to show distant matches and how the breakdown is. I still agree with Ann Turner that we are dealing with "small fuzzy boundaries" so not a real deal breaker.

Here is my v4 kit to my mother's on Gedmatch-Total cm-3547 largest cm-214.5
My FTDNA kit to her on Gedmatch- Total cm-3571.7 largest cm-209.3

On FTDNA- 3377.33 total cm largest block-267.21
On 23andme-3719 cM total largest block-284 cM

In general Gedmatch shows everything to be larger for the most part. Hopefully this is not so hard for everyone to look at.
I too have noticed differences in start-end positions. I attribute it to GEDmatch looking at the non-matching SNPs and seeing if it can make a matching block larger, by adding an adjacent small matching block and still not exceed a maximum of novel mutations in the combined block.
In other words I have seen some surprises when I did one-to-one without chromosome view and then latter with it; matches match better than I thought.
I was coming up with a shorter segment than my uncle to a match on Chr9 at FamilyTreeDNA. The cousin put her kit up on GEDmatch and now I can see that the difference comes down to the test detecting one novel SNP compared to him. Since the new small (IBD) segment was under 1cM FTDNA didn't report it, but GEDmatch --in chromosome view-- displays it plain as day as a half-match but below the 7cM threshold to be reported in the chart as a standalone segment, and getting a blue line under it in chromosome view. Since I'm a generation downstream from the match, compared to him, it makes since that a novel mutations would've occured.
Getting a little more to your observation, I think that in a case of a 3 half-identical adjoining segments divided with 2 non-adjacent SNPs, I think GEDmatch looks at the length of the segments and reports the middle segment and whichever adjoining segment is longer. Because this is not true below a certain minimum length (<<10cM) and over a certain maximum (like in the case of a whole chromosome) I'll state that I'm thinking of say a combined segment between 10 and 35cM.
I guess fuzziness of boundaries explains why 23andme rounds base positions to the nearest million.
Also, I have looked at the admix difference between matches with GEDmatch admix utils chromosome view where one party is 100% European and the other of mixed origins. There is fuzziness at either end well into the matching segment for the second party compared to the first. Because of that --instability-- I do not trust small (cM) scale admix "blips." I'd rather see a pattern lasting over many cM's, before I'd consider an admix component noteworthy --I've wondered if 8cM wouldn't be a better minimum threshold before toting up a component v. a chromosome.
Later,
still Befuddled.
dp

vettor
01-09-2015, 08:30 PM
Could you explain what you are saying in different words?

As per the ralph and coop paper............all national/regional populaces will give different results.
To conclude , I see a value in the program, and so does felix with his indian populace, you, being a central american might have different numbers which do not match.

Wolfie
01-09-2015, 09:56 PM
David, my half-sister and I match around 1800 cM, and are shown as 1 1/2 generations apart.

parasar
01-09-2015, 10:28 PM
As per the ralph and coop paper............all national/regional populaces will give different results.
To conclude , I see a value in the program, and so does felix with his indian populace, you, being a central american might have different numbers which do not match.

Europe north of the alpine chain must have been very minimally populated not that long back with much of its population growth occurring in the past 2000 years, especially in the eastern parts, and accelerating in the colonial and industrial age. The Mediterranean regions on the other hand must have sustained the vast bulk of European population before 2000ybp. Italy especially looks like a region with long term high populations.

Wolfie
01-09-2015, 11:57 PM
David, my half-sister and I match around 1800 cM, and are shown as 1 1/2 generations apart.

felix
01-10-2015, 03:20 AM
A population explosion in the past 500 years does not mean that a person from Ireland and a person from Yugoslavia (I'm using a hypothetical for the countries of origin) have a common ancestor in the past 500 years. Maybe in the past 2,000 or 4,000 years but not in the past 500 years. Ireland and Yugoslavia did not have colonies. Yugoslavs didn't migrate to Ireland and is it is extremely doubtful that Irish emigrated to Yugoslavia. If they both have trees going back at least 5 generations without any hint of either person have ancestry from the other country it is extremely doubtful that they have a common ancestry in the past 500 years and less doubtful that it just so happens that there is enough homozygous DNA from a common ancestor from 500 years ago to cause the calculator to think it is from a 5th cousin marriage. That's just ridiculous.

That is one huge IF by the way.

Yugoslavs don't had to migrate to Ireland for them to have cousins with each other. Two German brothers, one settling and mixing with Yugoslavs and another with Ireland is enough for a connection. Or Two 2nd cousins from Europe, one having a slave girl from Australia aboriginal and another traveled to New world have descendants will make descendants from Amerindians have 4th cousins with Aboriginals even though Amerindians never traveled to Australia. A Bantu raping a pygmy, and one of the descendants of this Bantu is captured as a slave girl to a European man, makes pygmies as cousins to European relatives back in Europe who had never set foot outside Europe. A king having lots of concubines and giving his daughter to another neighboring king who also has concubines will make the descendants of both concubines as cousins, even through both population never actually mixed.

Until you can account for 1 million ancestors 20 generations back (25 per generation) for the past 500 years, you can never assume the way you want. Just 1 out of 1 million ancestor having a relation with another population is enough for a person to be a cousin for another population.

Most of these facts are not recorded in pedigree or never mentioned in so called 'known relationships'. Population mixed a lot more than you think. So, don't blame me for having RoH in people's DNA.

ArmandoR1b
01-10-2015, 03:22 AM
As per the ralph and coop paper............all national/regional populaces will give different results.
To conclude , I see a value in the program, and so does felix with his indian populace, you, being a central american might have different numbers which do not match.

You didn't reword you original statement.

The Ralph and Coop paper does not prove that two people from completely different parts of a continent definitely have a common ancestor in the past 500 and even less so that there is enough homozygous DNA to make the program think that 80% of the people that use the program are 5th cousins.

You see a value because you aren't using logic. You are also making assumptions. I am not from Central America and some of the kits I am using in that program aren't from any colonial locations.

Apart from that, Felix has already admitted to and updated the page that has the calculator to state "The accuracy of the tool diminishes beyond 3rd cousin marriages". http://www.y-str.org/2014/12/pedigree-collapse-calculator.html Since the accuracy diminishes past 3rd cousin marriages the tool is no good for people that know for a fact their parents aren't 3rd cousins.

felix
01-10-2015, 03:40 AM
Apart from that, Felix has already admitted to and updated the page that has the calculator to state "The accuracy of the tool diminishes beyond 3rd cousin marriages". http://www.y-str.org/2014/12/pedigree-collapse-calculator.html Since the accuracy diminishes past 3rd cousin marriages the tool is no good for people that know for a fact their parents aren't 3rd cousins.

Yes, I am always happy to receive feedback and update the tools to provide better results. I also parameterized everything and the message will tell you the number of RoH segments.

I must add, it is never possible to have perfect calculators without taking a lot of factors into account, not just pedigree collapse but also twins, identical twins, triplets, NPE, duplicated ancestors etc. When the accuracy decreases beyond 3rd cousin, these unknowns play a heavy role in determining the percentage value, which could actually vary significantly from the truth. The message box is just the best possible guess used a guide. I never know why someone would take it too seriously.

ArmandoR1b
01-10-2015, 03:15 PM
Yugoslavs don't had to migrate to Ireland for them to have cousins with each other. Two German brothers, one settling and mixing with Yugoslavs and another with Ireland is enough for a connection. Or Two 2nd cousins from Europe, one having a slave girl from Australia aboriginal and another traveled to New world have descendants will make descendants from Amerindians have 4th cousins with Aboriginals even though Amerindians never traveled to Australia. A Bantu raping a pygmy, and one of the descendants of this Bantu is captured as a slave girl to a European man, makes pygmies as cousins to European relatives back in Europe who had never set foot outside Europe. A king having lots of concubines and giving his daughter to another neighboring king who also has concubines will make the descendants of both concubines as cousins, even through both population never actually mixed.

Until you can account for 1 million ancestors 20 generations back (25 per generation) for the past 500 years, you can never assume the way you want. Just 1 out of 1 million ancestor having a relation with another population is enough for a person to be a cousin for another population.

Most of these facts are not recorded in pedigree or never mentioned in so called 'known relationships'. Population mixed a lot more than you think. So, don't blame me for having RoH in people's DNA.

You are using the exact opposite of Occam's razor. You are trying to make the least likely cause to be the most likely cause. You are using exceptions and not rules. People have only recently been immigrating to Ireland. In the past there was no reason for anyone to immigrate there. Additionally, the odds of what you described happening plus the descendants marrying each other plus one of them using the calculator and seeing an amount of homozygosity that is equal to 5th cousin marriage is even more unlikely.

The Coop and Ralph paper supports what I am saying - "In most cases, only pairs within the same population are likely to share genetic common ancestors within the last 500 years." http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001555 If a person with ancestry in one country is most likely to only share ancestry with another person in the past 500 years marries a person from a distant country that is most likely to only share ancestry with someone from their own country then the likelihood that two people that do share ancestry from both of those countries is extremely low. For that to also happen to a large number of the users of the program, which is what was happening, is even more unlikely. The odds are completely against your hypothesis.

The accepted years per generation is 30 and not 25 and most people's genealogies that I can trace back to the 16th century are on average 30 years per generation. Going back 500 years at 30 years per generation is 16.7 generations. Rounding up to 17 generations that is only 131,072 ancestors. You are the one assuming what you want. Just because there have been instances that have never been recorded in a pedigree doesn't mean it happens often. The Ralph and Coop paper actually shows that people didn't mix as often as you think.

I am not blaming you for people having RoH I am refuting your statement that it must be due to common ancestry in the past 500 years. Something else you should take into consideration is that you aren't using parent-child trios and you aren't phasing the data to remove false positives so the calculator's result is much higher than reality.

Furthermore, Volume 15 | Number 9 | September 2013 | Genetics in medicine in a letter to the editor Noah A. Rosenberg, PhD, Trevor J. Pemberton, DPhil, Jun Z. Li, PhD and John W. Belmont, MD, PhD stated that RoH needs to be calculated differently for different populations. rosenberglab.stanford.edu/papers/RosenbergEtAl2013-GenetMed.pdf

ArmandoR1b
01-10-2015, 03:21 PM
Yes, I am always happy to receive feedback and update the tools to provide better results. I also parameterized everything and the message will tell you the number of RoH segments.

I must add, it is never possible to have perfect calculators without taking a lot of factors into account, not just pedigree collapse but also twins, identical twins, triplets, NPE, duplicated ancestors etc. When the accuracy decreases beyond 3rd cousin, these unknowns play a heavy role in determining the percentage value, which could actually vary significantly from the truth. The message box is just the best possible guess used a guide. I never know why someone would take it too seriously.

Because some tools and results can be taken seriously. If your tools are for just for laughs then that should be stated in the tool. Not everyone is looking for mindless entertainment.

vettor
01-10-2015, 04:51 PM
You didn't reword you original statement.

The Ralph and Coop paper does not prove that two people from completely different parts of a continent definitely have a common ancestor in the past 500 and even less so that there is enough homozygous DNA to make the program think that 80% of the people that use the program are 5th cousins.

You see a value because you aren't using logic. You are also making assumptions. I am not from Central America and some of the kits I am using in that program aren't from any colonial locations.

Apart from that, Felix has already admitted to and updated the page that has the calculator to state "The accuracy of the tool diminishes beyond 3rd cousin marriages". http://www.y-str.org/2014/12/pedigree-collapse-calculator.html Since the accuracy diminishes past 3rd cousin marriages the tool is no good for people that know for a fact their parents aren't 3rd cousins.

Ok, I reran the new test and went from 1.68% to 0.46%
But I still have marriages cerificates indicating a union between my paternal family tree side with my maternal family tree side circa 1800. Since I calculate a generation at 31 years ( as stated before) , it will put the union between 4th and 5th cousins.

I await a response of another match ( as I wrote to the person yesterday ) to get details and will find another paper/certificate trail and advise you

ArmandoR1b
01-10-2015, 06:30 PM
Ok, I reran the new test and went from 1.68% to 0.46%
But I still have marriages cerificates indicating a union between my paternal family tree side with my maternal family tree side circa 1800. Since I calculate a generation at 31 years ( as stated before) , it will put the union between 4th and 5th cousins.

I await a response of another match ( as I wrote to the person yesterday ) to get details and will find another paper/certificate trail and advise you

I also noticed a difference between the old calculator and the new calculator. I also increased the SNP count to 200 since David Pike has that as the default in his utility and that reduced the homozygosity even more. The pop-up after the calculator is done no longer happens with the kits either. That in-your-face pop-up was a bit annoying especially since it was incorrect.

I don't doubt your statement and your documentation that your father and your mother were related. What I was sure of is that the parents in the kits I ran through the kit aren't related in the past 6 generations. If a program tells me that the parents of an individual are related in the same degree as someone that does have a documented marriage between a recent ancestor in the paternal line and a recent ancestor in their maternal line then the program can't be relied upon even if it works for the person in the latter situation. Since the program is no longer telling me the kits are related within 6 generations (5th cousins) I no longer have that problem. The argument evolved into whether or not they have a common ancestor in the past 500 years and the Ralph and Coop study supports that they don't.

As I was typing this out I remembered that I have a kit of a person whose parents were 2nd cousins 1R and also related within 10 generations through several other ancestors. This person's parents knew that they were related even before the marriage and the documentation supports that. At 150 SNPs the person has 0.91% RoH and the pop-up shows up stating that they may be 5th cousins. So the program isn't even reliable when the parents actually are closer than 3rd cousins.

vettor
01-10-2015, 08:32 PM
I don't doubt your statement and your documentation that your father and your mother were related. What I was sure of is that the parents in the kits I ran through the kit aren't related in the past 6 generations.

for clarity..........my mother and father are not related, what is related are families in my paternal side of the tree, with families on the maternal side of the tree.

vettor
01-10-2015, 09:25 PM
for clarity..........my mother and father are not related, what is related are families in my paternal side of the tree, with families on the maternal side of the tree.

edit-
it is a great, great, great grandmother b.1771 on my maternal side ............married into my paternal side

felix
01-11-2015, 12:38 AM
Because some tools and results can be taken seriously. If your tools are for just for laughs then that should be stated in the tool. Not everyone is looking for mindless entertainment.

Shouldn't FTDNA's 2nd to 3rd cousin taken too seriously when it does not provide accurate result always?
Shouldn't 23andMe's Neanderthal calc taken too seriously for not using neanderthal genome but using Africans and Europeans to calc the results?

I think you purposefully want to put me down for no reason.

felix
01-11-2015, 12:57 AM
You are using the exact opposite of Occam's razor. You are trying to make the least likely cause to be the most likely cause. You are using exceptions and not rules. People have only recently been immigrating to Ireland. In the past there was no reason for anyone to immigrate there. Additionally, the odds of what you described happening plus the descendants marrying each other plus one of them using the calculator and seeing an amount of homozygosity that is equal to 5th cousin marriage is even more unlikely.

The Coop and Ralph paper supports what I am saying - "In most cases, only pairs within the same population are likely to share genetic common ancestors within the last 500 years." http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001555 If a person with ancestry in one country is most likely to only share ancestry with another person in the past 500 years marries a person from a distant country that is most likely to only share ancestry with someone from their own country then the likelihood that two people that do share ancestry from both of those countries is extremely low. For that to also happen to a large number of the users of the program, which is what was happening, is even more unlikely. The odds are completely against your hypothesis.

The accepted years per generation is 30 and not 25 and most people's genealogies that I can trace back to the 16th century are on average 30 years per generation. Going back 500 years at 30 years per generation is 16.7 generations. Rounding up to 17 generations that is only 131,072 ancestors. You are the one assuming what you want. Just because there have been instances that have never been recorded in a pedigree doesn't mean it happens often. The Ralph and Coop paper actually shows that people didn't mix as often as you think.

I am not blaming you for people having RoH I am refuting your statement that it must be due to common ancestry in the past 500 years. Something else you should take into consideration is that you aren't using parent-child trios and you aren't phasing the data to remove false positives so the calculator's result is much higher than reality.

Furthermore, Volume 15 | Number 9 | September 2013 | Genetics in medicine in a letter to the editor Noah A. Rosenberg, PhD, Trevor J. Pemberton, DPhil, Jun Z. Li, PhD and John W. Belmont, MD, PhD stated that RoH needs to be calculated differently for different populations. rosenberglab.stanford.edu/papers/RosenbergEtAl2013-GenetMed.pdf

Since you showed that pairs within the same population in Europe are likely to share genetic common ancestors within the last 500 years, then it certainly makes sense when 1 million Europeans who came to Asia in 200 years (1600 to 1800 alone through VoC) and had families with locals made the locals cousins with Europeans - Unless you are so narrow minded and work desperately to exclude those genetic cousins.

ArmandoR1b
01-14-2015, 06:18 PM
Since you showed that pairs within the same population in Europe are likely to share genetic common ancestors within the last 500 years, then it certainly makes sense when 1 million Europeans who came to Asia in 200 years (1600 to 1800 alone through VoC) and had families with locals made the locals cousins with Europeans - Unless you are so narrow minded and work desperately to exclude those genetic cousins.

You are being narrow minded because you think that is the situation with the kits I have used in your calculator. You are being narrow minded when you do not realize that a person from one country that is most likely to only match other countrymen in the past 500 years marries a person from another distant country that is most likely to only match another countryman in the past 500 years their offspring are very unlikely to have as much homozygous DNA as a person with parents that are 5th cousins when some of them have trees going back to the 16th century in countries that have nothing to do with the other. The parents and relatives of these people's parents do not have any matches whatsoever with people from the country of the other parent. The people that show to have homozygous DNA in your calculator do not show their parents to be related in the Gedmatch "Are your parents related?" calculator while the one that does have parents that were related did show to have parents that were related in the Gedmatch calculator.

dp
01-15-2015, 05:20 PM
Because how a persons parents are related is related to their amount of IBD sharing I believe readers of this thread would find the following abstracts I made from a science paper of interest. Please go to the following post on European IBD segment sharing (http://www.anthrogenica.com/showthread.php?3132-Blurring-the-line-between-genealogical-genetics-and-population-admixture&p=64800&viewfull=1#post64800)
dp :-)

dp
01-16-2015, 03:37 PM
Below is a comparison of my mom's matching segments on Chr 20-22 compared to my own, in the cases when we matched the same person's on the same segments. It is to show how a child's segments can be longer at an end, or both than a parents.
Click the spoiler button to view:

http://www.anthrogenica.com/attachment.php?attachmentid=3393&d=1421422288

Basepair Color Coding
Green: My endpoint extends further than hers. Yellow: we have the same endpoint. Red(dish): my endpoint is less than hers.
Bright Green: segments where I match either more than her on both ends, or one end is the same as hers, but the other is longer. This was done because direct cM comparisons can be made without extra calculation to see the amount of additional segment length in a child - parent.
dp :-)

PS: in a PM with Felix the topic came up so credit where due.