Page 8 of 8 FirstFirst ... 678
Results 71 to 79 of 79

Thread: For anyone out there who doesn't believe Timber is a bad idea

  1. #71
    Registered Users
    Posts
    1,535
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    Quote Originally Posted by jadegreg View Post
    I continue to encounter similar issues as yourself Geebee. It really is quite laughable

    Do we know if this consequently impacts our genetic communities (GCs)? My sister and I share identical matching for a couple of 4th cousins, across 3 segments, who get Greater London and SC England for their communities. However, Timber, has massively reduced my cM match on a couple of segments. Consequently, I beleive, I only get Greater London as GC, whilst my sister gets SC England community.

    Your thoughts?
    It's difficult to say how much Timber is responsible for differences in Genetic Communities between siblings. It's possible that it has some effect, but I suspect not as great an effect as the simple fact that siblings only share about half their DNA if they're full siblings, and a quarter if they're half siblings.

    If you think of your GC's as the result of DNA inherited from your grandparents, each sibling's inheritance from any grandparent can be considerable.

    To illustrate, my father and five of his grandchildren -- and one great grandchild -- all tested at 23andMe. While we generally like to think that everyone gets about a quarter of their DNA from each grandparent, it can actually be several percent higher or several percent lower. For example, one grandchild inherited about 28% from my father, while another inherited only 18%. The flip side of this is that the one who inherited 28% from my father must have inherited only 22% from my mother, since the inheritance from a pair of grandparents on the same side will add up to 50%. So the grandchild who inherited 18% from my father also inherited 32% from my mother.

    These two happen to be cousins, but even siblings can have a large difference in their inheritance from a shared grandparent. The sister of the grandson who inherited 28% from my father only inherited 20% from my father, and therefore 30% from my mother.

    You can see this "differential inheritance" better when you can compare sibling results from companies -- like 23andMe -- that provide a chromosome browser. Ancestry's lack of one is really one of the biggest downsides to the company. They try to make up for it with other tools, but for some things (such as triangulation) there really is no substitute for a chromosome browser.

    Another thing 23andMe does that few other companies do is they show you where on each chromosome a given ancestry appears. This is great when you're comparing siblings, since you can see whether a segment that should be matching really is. For example, if Ancestry Comparison says one of my siblings and I have different "ethnicities" in a region where we have full matching (matching on both chromosomes), I know it has to be wrong for at least one of us. Fortunately, this doesn't happen often. More frequently, the calls are consistent.

    With just percentages, you can't make that sort of comparison.
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  2. The Following 3 Users Say Thank You to geebee For This Useful Post:

     jadegreg (02-15-2021),  Riverman (05-09-2021),  Rufus191 (02-25-2021)

  3. #72
    Registered Users
    Posts
    1,535
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    So, I recently encountered another match too big for Timber to mess with: 145 cM across 9 segments, with a longest segment of 41 cM. Now, given these numbers it really wouldn't be surprising for my daughter to share something like 87 cM with this person, right?

    Well, although that's the unweighted sharing my daughter received, Timber obviously "thought" that it was too matchy, and so even though the sharing was a mere 3 cM less than the "safe" amount of 90 cM or above, it adjusted this to just 60 cM. Or in other words, it discounted the shared cM to less than 70% of the original amount.

    Of course, if there were a chromosome browser I could actually look and see if if the 3 segments my daughter shares with the match are located in the same places as 3 of my 9 shared segments. At least one very likely is, since my daughter has exactly the same length of "longest segment" that I do: 41 cM.

    They almost certainly won't do this, but the only way that it would really be possible to test the reliability of Timber would be to use it in conjunction with a chromosome browser. I am pretty sure that this would confirm what I've been seeing all along, that even if Timber might sometimes be right, it's too often wrong to be worth using.

    I'd much rather be "misled" by having some matches more distant matches appear close than by having close matches appear more distant.
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  4. The Following User Says Thank You to geebee For This Useful Post:

     Saetro (06-13-2021)

  5. #73
    Registered Users
    Posts
    62
    Sex
    Location
    Tacoma, Washington, USA
    Ethnicity
    Afro,Euro,Amerind
    Nationality
    American
    Y-DNA (P)
    E-M85
    mtDNA (M)
    J1c14

    United States of America
    Quote Originally Posted by geebee View Post
    It's difficult to say how much Timber is responsible for differences in Genetic Communities between siblings. It's possible that it has some effect, but I suspect not as great an effect as the simple fact that siblings only share about half their DNA if they're full siblings, and a quarter if they're half siblings.

    If you think of your GC's as the result of DNA inherited from your grandparents, each sibling's inheritance from any grandparent can be considerable.

    To illustrate, my father and five of his grandchildren -- and one great grandchild -- all tested at 23andMe. While we generally like to think that everyone gets about a quarter of their DNA from each grandparent, it can actually be several percent higher or several percent lower. For example, one grandchild inherited about 28% from my father, while another inherited only 18%. The flip side of this is that the one who inherited 28% from my father must have inherited only 22% from my mother, since the inheritance from a pair of grandparents on the same side will add up to 50%. So the grandchild who inherited 18% from my father also inherited 32% from my mother.

    These two happen to be cousins, but even siblings can have a large difference in their inheritance from a shared grandparent. The sister of the grandson who inherited 28% from my father only inherited 20% from my father, and therefore 30% from my mother.

    You can see this "differential inheritance" better when you can compare sibling results from companies -- like 23andMe -- that provide a chromosome browser. Ancestry's lack of one is really one of the biggest downsides to the company. They try to make up for it with other tools, but for some things (such as triangulation) there really is no substitute for a chromosome browser.

    Another thing 23andMe does that few other companies do is they show you where on each chromosome a given ancestry appears. This is great when you're comparing siblings, since you can see whether a segment that should be matching really is. For example, if Ancestry Comparison says one of my siblings and I have different "ethnicities" in a region where we have full matching (matching on both chromosomes), I know it has to be wrong for at least one of us. Fortunately, this doesn't happen often. More frequently, the calls are consistent.

    With just percentages, you can't make that sort of comparison.
    With me being a multiracial person, knowing that matching segment location is European or SubSaharan African is very helpful.
    Chromosome browser helped me learn where some of my paternal European segments come from Acadians.
    My father was an African American born and raised in New Orleans. His paternal grandmother was born and raised in Southwestern Louisiana, and she had some Acadian ancestry. My father's maternal half sister matched me on my paternal European segments where I matched people with Acadian ancestry.
    Even though she's African American, the longest segment that she and I share is my whole entire European segment on paternal Chromosome 5. It's about 90 cM.


    In regards to the topic of this thread, I don't like Timber. I get matches at other sites that don't show up at AncestryDNA.
    My paternal matches with Acadian ancestry that match me on other sites don't show up AncestryDNA.
    There are people with Acadian ancestry that match both my paternal aunt and me in the same location with around the same length, and they show up as much smaller match to me and no match to my paternal aunt. That even goes for segments that are around 30 cM.
    Last edited by Glaucus; 06-18-2021 at 06:26 AM.
    paternal ancestry is African American with English and Acadian
    maternal ancestry is Cape Verdean, Ashkenazi Jewish (from Romania and Latvia), Puerto Rican, Madeiran, Azorean, English, Scottish, German, Irish, Swiss, Welsh, Dutch, Frisian


    Eurogenes K13
    Sub_Saharan 47.34%
    North_Atlantic 18.75%
    West_Med 9.90%
    East_Med 6.70%
    Baltic 4.70%
    Northeast_African 3.60%
    West_Asian 2.73%
    Amerindian 2.38%
    Red_Sea 1.86%
    East_Asian 1.20%
    Siberian 0.64%
    Oceanian 0.11%
    South Asian 0.10%

  6. The Following User Says Thank You to Glaucus For This Useful Post:

     geebee (06-18-2021)

  7. #74
    Registered Users
    Posts
    1,535
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    Here's how rigid Ancestry is in its application of Timber. I just found a new match, whom I'll call "PWB". Ancestry says we share 78 cM across 6 segments. But the unweighted amount is given as 90 cM, with a longest segment of 21 cM.

    Wait! I thought any match that shares 90 cM or more was "safe" from Timber? Ah, but what Ancestry really means is 90.0 cM or above, and presumably my sharing with PWB actually only rounds to 90 cM. That means the true amount is likely between 89.5 and 89.9 cM. So for want of 0.5 cM or less, Timber was free to chop about 12 cM off the match -- or some 13%.
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  8. The Following 4 Users Say Thank You to geebee For This Useful Post:

     digital_noise (06-19-2021),  Dts (06-19-2021),  Glaucus (06-21-2021),  Rufus191 (06-22-2021)

  9. #75
    Registered Users
    Posts
    1,535
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    Yet another example of how Timber is too quick to chop:

    The granddaughter of one of my 1st cousins -- so, my 1st cousin twice removed -- and I share 237 cM across 12 segments, with a longest segment of 70. Since this total is way over 90 cM, it means Timber doesn't mess with it.

    The same person's sharing with my daughter, however, is given as just 46 cM across 4 segments. The total shared amount is actually only about a fifth of mine. While this is possible, it turns out that the unweighted amount is 81 cM, with a longest shared segment of 50 cM. That means that before Timber, the three remaining segments must have totaled 31 cM. So Timber has reduced matching by 38% when that clearly isn't necessary.

    If I share 237 cM with someone, there's no reason to suppose my daughter couldn't share 81 cM with the same person. In fact, it's somewhat more likely than that she would share just 46 cM -- 4 cM less than the longest segment all by itself!

    Of course, if there were a chromosome browser, I could easily check to see how the sharing compared for all three of us. Ancestry, however, claims "privacy" concerns as the reason for not providing a chromosome browser. That may be true -- it is a legitimate concern.

    It is also convenient, since ultimately a chromosome browser is the only way to determine just how often Timber's chopping is not only unnecessary, but a hindrance.
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  10. The Following User Says Thank You to geebee For This Useful Post:

     Glaucus (06-24-2021)

  11. #76
    Registered Users
    Posts
    1,535
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    Here's a nice one.

    "JC" shares 26 cM in a single segment with me -- so claims Timber. The trouble is, the "longest segment" is given as 43 cM. So Timber is telling me that 17 cM of this segment is simply "excess matching".

    But that isn't the only problem. My daughter, according to Ancestry, shares 36 cM with JC, also in a single segment. The unweighted amount is 42 cM -- meaning that I basically passed the entire segment on to her, intact. Not unusual for a segment of this size. Obviously, Timber is still downgrading cM here, but less for my daughter than for me by a fairly large margin.

    Look at this important difference, though, caused entirely by Timber: judging by the unweighted amounts, it's simply the same segment escaping a crossover and being passed on intact, but by Timber's calculation it's 10 cM more being passed on to my daughter than I supposedly have.

    Further, there's really no way to be sure how close or distant this match is -- he might be as close as a 3rd cousin, though he can't be a 3rd cousin to both of us. At a guess he might be something like a 3rd cousin or 3rd cousin once removed to me, which would likely make him a 3rd cousin or 3rd cousin once removed to my daughter. Or, more distant is possible.

    Without a chromosome browser (or a tree on his side), there's no way to tell. But to me that means the reliance on Timber is even less justified by Ancestry. (Since it's impossible to check.)
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  12. #77
    Registered Users
    Posts
    1,535
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    One thing I've found useful is Ancestry's new feature allowing us to choose a side and relationship. I've been working through this in conjunction with the "Common Ancestors" filter.

    In the process, I've encountered a lot of matches like this one:

    "LB" shares 14.9 cM with me -- according to Timber -- in a single segment. However, unweighted sharing is 47 cM. That means that Timber "thinks" that 2/3 of this single segment is really "excess matching". Based on Timber's 15 cM, the first choice offered in "Relationship Assignment" is 3rd Cousin once removed, which Ancestry says has a 22% probability. That isn't high, but no other relationship has any higher of a probability.

    I can, however, select "Some more possible relationships" and choose "3rd cousin" -- which is the correct relationship according to "Common Ancestors". Ancestry shows this as having only a 14% probability. But that's based on Timber's downgraded cM amount.

    I used the "Shared DNA" filter and selected a range of 47-47 cM -- which basically just limited the matches show to those who were at least 47 cM but less than 48 cM. There was only one CA match that fit the criteria, and he happens to be another 3rd cousin. Using the "Edit function" for side and range, the first choice offered by Ancestry is the same one I selected, with a 35% probability according to Ancestry.

    My point is, "3rd" cousin -- the actual relationship -- would have been Ancestry's "1st pick" for my relationship with LB, except for Timber's downward adjustment of cM.

    Now, has this always been the case with the matches I've looked at so far? No, but it's happened fairly often. And in each case, relying on Timber would actually have made the relationship look more distant -- the exact opposite of what Ancestry says is Timber's purpose.

    I've said it before, but I'd much rather see a match that seems close enough to be worth evaluating it, only to find that it didn't "pan out", than to have Ancestry tell me to give the match a lower priority. At least now they do allow me to see unweighted sharing and the longest segment, but only after I take a closer look at the match.

    This unweighted cM and longest segment should be included in the 2nd column of the match list. It's more worthwhile than "percent", especially when they're already listing "cM", so if it's a space issue unweighted cM and longest segment could replace that part. In addition, it should be possible to reorder the match list -- perhaps by "Relationship", "shared cM" (Timber), "unweighted cM", and "longest segment".
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  13. #78
    Registered Users
    Posts
    1,535
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    Here's one reason why it's very unhelpful for Ancestry not to have a chromosome browser.

    One of my matches, "PJF", is reported by Ancestry to share 28.3 cM with me across 2 DNA segments. However, unweighted sharing is 58 cM, and longest segment is 45 cM. So "Timber" has chopped our matching almost in half.

    Now, the relationship between us is known: 3rd cousin once removed. Or, to put it a little more precisely, PJF's mother and I were 3rd cousins.

    It turns out, though, that PJF has tested at MyHeritage -- which does have a chromosome browser. And MyHeritage does not say we match on two segments, but only one -- on chromosome 12. MyHeritage also estimates this to be just one segment, not two, and 63.7 cM in length. This likely means that there were at least a few SNPs between the two sections that MyHeritage sees as a single segment, which at Ancestry were discordant. Normally, a single discordance would not cause a "break" to be called, but two or more SNPs within a certain distance might.

    In either case, it is clear that the "two" segments are on the same chromosome, and it's likely that at some point they were part of a single segment -- on both sides. That's one indicator of fairly recent common ancestry.

    I will note that at MyHeritage, the shared segment spans the centromere, and centromeres are often regions that are not terribly SNP-dense. In any case, though, MyHeritage does report a rather large amount of matching (or technically, half matching) SNPs: 32,256.

    There's such a difference between claiming that a match is only 28.3 cM long, spread across 2 difference segments that might not even be on the same chromosome, and claiming that a match is actually 63.7 cM in length. A segment of this length on chromosome 12 covers over two-thirds of the entire chromosome*, yet Ancestry has the audacity to claim "excess sharing" requires to report sharing as less than half of that.

    *Using chromosome length as given for GEDmatch at ISOGG, since there is not a calculation specifically for MyHeritage. https://isogg.org/wiki/CentiMorgan
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  14. The Following User Says Thank You to geebee For This Useful Post:

     peloponnesian (06-30-2021)

  15. #79
    Registered Users
    Posts
    1,535
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    One of the problems I have with Ancestry's use of "Timber" is that there seems to be a presumption that if the amount of shared DNA seems to be excessive -- because of multiple distant shared ancestors -- that this is necessarily not useful information.

    However, just because two people may have several distant shared ancestors instead of fewer more, recent shared ancestors does not automatically make these ancestral paths untraceable.

    For example, I have relative -- "DAC" -- who tested at both 23andMe and Ancestry. According to 23andMe, we share 90 cM across 6 segments. But, since one of these segments -- 19 cM in length -- is on the X chromosome, Ancestry automatically excludes it. The total length of the remaining segments is deeply discounted by Ancestry, which reports our sharing as just 21 cM across 5 segments.

    However, the unweighted sharing is given as 70 cM, and the longest segment all by itself is 35 cM. Plus, three of the shared segments lie on the same chromosome and therefore might at some time have been part of a single, longer segment.

    Yet in actuality, DAC and I do not appear to be close relatives. In fact, based on our respective trees at Ancestry, we are 5th cousins once removed. We are also 6th cousins once removed. But using my genealogy software, it turns out that our 6th cousin relation is actually five-fold. That is, we're 5th cousins once removed on one path, and 6th cousins by five different paths. Four of these trace back to the same set of 5th great grandparents, although on my side I descend from this couple only twice.

    Now, not everyone necessarily has this sort of information, but some people do. And while this may mean that I indeed have some "excess matching" due to multiple shared ancestry, it isn't really surprising that I may have "excess DNA" from certain ancestors -- and not always different ancestors.

    Further, when I look at the 5th great grandparents may match and I share by four different pathways, I see that Ancestry reports that the number of ThruLines descendants of this couple that share DNA with me is 754 (although it appears that some of the offspring may actually be duplicates of each other).

    From a statistical standpoint, once a couple has five children, they have a pretty good chance of having passed on all their DNA to their offspring. With one child, each parent passes on half of his or her DNA. With the second child, on average half of DNA each child receives from a parent will be the same, and half will be different.

    With this particular set of ancestors, there are many more than five offspring -- even eliminating duplicates. Plus, those offspring largely each had numerous children, as did the children's children. Not only that, but in this part of my family there were often marriages between cousins -- though not usually between 1st cousins.

    The point is, however, that in many instances the same DNA that might drop out in one generation (on a given line) could be reintroduced in another (from a different line).

    In the case of many of these ThruLines matches, the difference between Timber's reported cM and unweighted cM is not that great. In other instances, however, it is considerable. So what I wonder is, does this really reflect lots of very small segments from lots of distant shared ancestors -- that just happen to be located in the right places on both sides to look like fewer, larger shared segments -- or does it sometimes represent fewer, larger shared segments that managed to be transmitted a greater distance simply because there were more chances for them to do so?

    Because of Ancestry's lack of any sort of chromosome browser, there's no real way to investigate this. But here's what our sharing looks like in 23andMe:

    GJSB versus DAC.JPG

    Also, our shared ancestry includes a Native American 5th great grandfather -- who occupies more than one slot in our respective tree, due to endogamous marriages. Chromosome 6 includes some DNA which Ancestry Composition labels as "Indigenous American", although I'm unfortunately no longer able to see the ancestry painting for my match. But here you can see our respective percentages.

    Indigenous American - GJSB & DAC.JPG
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  16. The Following 3 Users Say Thank You to geebee For This Useful Post:

     Dts (07-18-2021),  Glaucus (07-12-2021),  Stefanie (07-12-2021)

Page 8 of 8 FirstFirst ... 678

Similar Threads

  1. This doesn't make sense
    By firemonkey in forum AncestryDNA
    Replies: 2
    Last Post: 01-29-2020, 06:52 AM
  2. AC/Admix on the X Chromosome (doesn't match Hap)
    By Unfounded in forum X Chromosome (X-DNA)
    Replies: 10
    Last Post: 09-12-2019, 01:36 PM
  3. So who has steppe and who doesn’t?
    By Censored in forum Southern
    Replies: 20
    Last Post: 06-28-2018, 02:21 PM
  4. Replies: 11
    Last Post: 08-05-2015, 02:17 AM
  5. Exercise Doesn't Help with Weight Loss?
    By Scarlet Ibis in forum Health and Fitness
    Replies: 20
    Last Post: 08-26-2012, 04:12 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •