PDA

View Full Version : Debunking Relative Finder for Ashkenazi ancestry



AJL
05-21-2014, 06:01 PM
A large number of customers of Ashkenazi descent have bought 23andme, both for health and ancestry reasons.

Here's why Relative Finder is useless for Ashkenazi ancestry.

In one entirely Ashkenazi person's profile, these surnames are most common among matches:

1. Cohen
2. Katz (variant of Cohen)
3. Levine
4. Goldberg
6. Friedman
7. Stern
8. Greenberg
9. Segal (variant of Levine)
10. Horowitz (variant of Levine, Segal)
11. Lifshitz
12. Finkelstein
13. Shapiro (variant of Cohen, Katz)
14. Schwartz
15. Epstein

Compare these to the most common Ashkenazi surnames of arrivals at Ellis Island in the late 19th century/early 20th, courtesy Yannay Spitzer (http://yannayspitzer.net/2012/07/24/most-common-jewish-names/):

1. Levin/Lewin (variant of 3 on list)
2. Kaplan (variant of 1,2,13 on list)
3. Goldberg (4 on list)
4. Katz (2 on list, also variant of Kaplan etc.)
6. Goldstein
7. Cohen (1 on list)
8. Schapiro (13 on list, often also a variant of Cohen etc.)
9. Epstein (15 on list)
10. Rosenberg
11. Friedman (6 on list)
12. Weinstein
13. Schneider
14. Feldman
15. Kagan (variant of 1,2,13 on list)

Ah well, the lists are almost the same. What does this mean?

In brief, almost all links are specific to a gene pool rather than a family. There is no way to discern actual ancestral ties via shared surnames for anything closer than about a second or third cousin, as one would expect since most of these surnames were independently adopted about 1770. In the same person's kit, there are 149 predicted second-to-third cousins among 3,065 DNA relatives, none of whom is a 2nd/3rd cousin by paper trail.

Let's also take a closer look at some of these surnames and how they correlate to haplogroups:

Horowitz: J1, Q1b, R1a1a, T
Levin, Levine: E1b1b1c1a, G2a4, G2c, J1, J1e, J2, J2a1b1, R1a1a, R2
Segal: R1a1a, R1b1b2
Cohen / Kaplan / Katz / Kagan: E1b1b1c1, E1b1b1c1a, G2a4, G2c, J1, J1e, J2, J2b2

Need I say more? :\

Táltos
05-21-2014, 07:54 PM
Interesting AJL. I have very distant Jewish ancestry, so obviously my top surname profile doesn't look like the one you have listed. The top Jewish surnames most common among my matches are:
Rosenberg at #2 spot
Cohen makes #11

Joe B
05-21-2014, 08:42 PM
Interesting AJL. I have very distant Jewish ancestry, so obviously my top surname profile doesn't look like the one you have listed. The top Jewish surnames most common among my matches are:
Rosenberg at #2 spot
Cohen makes #11
My Ashkenazi admixture is only 4.7% in standard or 6.1% in speculative. What seems like Jewish surnames dominate my DNA relatives, just as AJL was explaining. Schwartz is a common German surname by the way.


surname closest relative count enrichment
Levin 4th Cousin 8 103
Cohen Distant Cousin 10 76
Schwartz 4th Cousin 8 70
Katz 4th Cousin 5 52
Goldstein Distant Cousin 5 52
Levine 4th Cousin 5 50
Goldberg 4th Cousin 5 48
Brown 4th Cousin 6 5
Davis 4th Cousin 5 4
Smith 4th Cousin 5 0

AJL
05-21-2014, 09:36 PM
^ The "enrichment" does not seem to take this into account adequately.

evon
05-21-2014, 10:00 PM
The “Uniqueness” of Ashkenazi Jewish Ancestry is Important for Health

http://blog.23andme.com/ancestry/the-uniqueness-of-ashkenazi-jewish-ancestry-is-important-for-health/


I probably have distant Ashkenazi ancestry, but more recent Sephardi ancestry..Hoping the rumors that Myorigins will include a Sephardi cluster this summer are true..

The Ashkenazi Jewish names that we have among or relatives are Hartmann, Katz, Lippmann, Harris, Sanders, Rosenfeld, Levin, Shatcher, Rabinowitz, Rothstein, ect..But the most common is Hartmann i think, but its connected via Romani linage also, so its likely not directly Ashkenazi for my linage...

AJL
05-21-2014, 10:49 PM
Lippmann is sometimes a given name and sometimes a surname that may be related to Heller/Geller, as in:

http://kehilalinks.jewishgen.org/svisloch/descendants_of_yom_tov_lippman.htm

Perhaps in your case it's worth noting that one of his lines had descendants in Copenhagen as of the late 17th/early 18th century.

Táltos
05-21-2014, 11:43 PM
My Ashkenazi admixture is only 4.7% in standard or 6.1% in speculative. What seems like Jewish surnames dominate my DNA relatives, just as AJL was explaining. Schwartz is a common German surname by the way.


surname closest relative count enrichment
Levin 4th Cousin 8 103
Cohen Distant Cousin 10 76
Schwartz 4th Cousin 8 70
Katz 4th Cousin 5 52
Goldstein Distant Cousin 5 52
Levine 4th Cousin 5 50
Goldberg 4th Cousin 5 48
Brown 4th Cousin 6 5
Davis 4th Cousin 5 4
Smith 4th Cousin 5 0

Oh my goodnes Joe your Ashkenazi is a lot compared to mine there! I have 0.9 in all three modes. My daughter has 0.9 in conservative, and standard, but gets up to 1.0 in Speculative. Her split view shows it is all from me. Now just how in all heck is myOrigins telling me I have 5% Jewish Diaspora? How much JD did myOrigins give you if you don't mind me asking?

Also the Rosenberg, and Cohen are predicted 4th cousin for me. I have one of the Cohen matches at FTDNA, they are predicted 3rd cousin there. :\ That one even made myOrigins for sharing there.

AJL
05-22-2014, 12:58 AM
Also the Rosenberg, and Cohen are predicted 4th cousin for me. I have one of the Cohen matches at FTDNA, they are predicted 3rd cousin there. :\ That one even made myOrigins for sharing there.

This will always be a problem with Ashkenazi : non-Ashkenazi matches. Do we assume the match is Ashkenazi then downgrade the proximity, or assume it isn't and upgrade it?

Perhaps a reasonable (though complicated) approach would be to paint parts of the chromosome, and see whether individuals match on an Ashkenazi-painted, East Euro–painted, etc., area.

This also gets complicated when dealing with two different endogamous populations. My Ashkenazi grandfather has a small section, 0.1% of his genome, at 23andme that's painted Finnish. I have no reason to doubt it's real, but since he also shares larger segments that date to the Spanish Expulsion in the 1490s, who's to say the Finnish link isn't from, say, the Volga River 1,500 years ago?

Joe B
05-22-2014, 01:38 AM
Oh my goodnes Joe your Ashkenazi is a lot compared to mine there! I have 0.9 in all three modes. My daughter has 0.9 in conservative, and standard, but gets up to 1.0 in Speculative. Her split view shows it is all from me. Now just how in all heck is myOrigins telling me I have 5% Jewish Diaspora? How much JD did myOrigins give you if you don't mind me asking?

Also the Rosenberg, and Cohen are predicted 4th cousin for me. I have one of the Cohen matches at FTDNA, they are predicted 3rd cousin there. :\ That one even made myOrigins for sharing there.
I didn't transfer my results over to FTDNA. If I recall from our mapping exercises we were doing right here at Anthrogenica, we were both plotting around the Lusatian Mountains between Germany and the Czech Republic for two population percentages from gedmatch. Not to mention a previous post.
Hey Joe B so far your results look closest to mine! B)
Here are mine again to compare.
EEF=55.67025
WHG=30.18771
ANE=14.14204
Does FTDNA's Jewish Diaspora have a different genetic definition than Ashkenazi at 23andme? That would explain your higher 5.0% Jewish Diaspora.

With my documented Irish and the experience of over sampling from the Isles with the Y-chromosome, I was expecting a few more Irish surnames.

AJL
05-22-2014, 02:06 AM
With my documented Irish and the experience of over sampling from the Isles with the Y-chromosome, I was expecting a few more Irish surnames.

The problem is that the cutoff for Relative Finder is a low amount, I think 7 cM, so your RF will be populated mainly by Ashkenazi matches if you have anything over a few percent Ashkenazi.

vettor
05-22-2014, 03:15 AM
The problem is that the cutoff for Relative Finder is a low amount, I think 7 cM, so your RF will be populated mainly by Ashkenazi matches if you have anything over a few percent Ashkenazi.

what is the % for "noise" ( of any ethnicity) in 23andme?

a year ago , I had 0% ashkenaki, then 6 monts later, I had 0.4% , then a month ago it was 0.8% and now I have 0.7% .........always only on Chr#6 ......its on the X Chr in exactly the same place as DougM marked me as south asia

Táltos
05-22-2014, 03:24 AM
This will always be a problem with Ashkenazi : non-Ashkenazi matches. Do we assume the match is Ashkenazi then downgrade the proximity, or assume it isn't and upgrade it?

Perhaps a reasonable (though complicated) approach would be to paint parts of the chromosome, and see whether individuals match on an Ashkenazi-painted, East Euro–painted, etc., area.

This also gets complicated when dealing with two different endogamous populations. My Ashkenazi grandfather has a small section, 0.1% of his genome, at 23andme that's painted Finnish. I have no reason to doubt it's real, but since he also shares larger segments that date to the Spanish Expulsion in the 1490s, who's to say the Finnish link isn't from, say, the Volga River 1,500 years ago?
Thank goodness 23andme does chromosome painting, and I can see they match me on the same spot of my chromosome 7 that is painted Ashkenazi. My Jewish matches I can see that same chromosome for them is overwhelming Ashkenazi on top and bottom, but especially in the spot they match me.

Táltos
05-22-2014, 03:35 AM
I didn't transfer my results over to FTDNA. If I recall from our mapping exercises we were doing right here at Anthrogenica, we were both plotting around the Lusatian Mountains between Germany and the Czech Republic for two population percentages from gedmatch. Not to mention a previous post.
Does FTDNA's Jewish Diaspora have a different genetic definition than Ashkenazi at 23andme? That would explain your higher 5.0% Jewish Diaspora.

With my documented Irish and the experience of over sampling from the Isles with the Y-chromosome, I was expecting a few more Irish surnames.
Thanks Joe I see. Believe me you aren't missing much with their autosomal test LOL! Yes I do remember our maps had been very similar, as was our EEF, WHG, and ANE. :D

I just reread this over at FTDNA https://www.familytreedna.com/learn/ftdna/behind-myorigins/

This part caught my eye:

Rather than the history of your family, myOrigins attempts to place your genetic makeup in a broader framework of variation and relatedness which is the outcome of a sequence of separations and reunions.

And then it's goal:

The goal of myOrigins was to give geographically relevant answers to people about their deep ancestry.

Really FTDNA, not what I'm looking for. This is more for conversation at a cocktail party.

AJL
05-22-2014, 03:38 AM
what is the % for "noise" ( of any ethnicity) in 23andme?

a year ago , I had 0% ashkenaki, then 6 monts later, I had 0.4% , then a month ago it was 0.8% and now I have 0.7% .........always only on Chr#6 ......its on the X Chr in exactly the same place as DougM marked me as south asia

McDonald says ~1% for his test. 23andme has more samples, so you could possibly go down a little from there, but I am still suspicious of some of these 0.1-0.2% things on 23andme. 0.6-0.7% seems fairly solid, but Chromosome 6 contains a large part of the immune system, so it may not be the best one to base traces of ancestry on.

vettor
05-22-2014, 06:05 AM
McDonald says ~1% for his test. 23andme has more samples, so you could possibly go down a little from there, but I am still suspicious of some of these 0.1-0.2% things on 23andme. 0.6-0.7% seems fairly solid, but Chromosome 6 contains a large part of the immune system, so it may not be the best one to base traces of ancestry on.

thanks

so the X chr which in 23andme has a small Ashenzaki marker which matches Ch6 in 23andme a a small Ashenzaki ( same numbers ) and matches exactly my DougM 's X chr as being South-Asian , can mean basically nothing in terms of finding any links to people?

Can you link anything about chr 6 being the immune sytem?

esev
05-22-2014, 12:04 PM
Hello all,
I have the weirdest result. First of all I am very admixed.
On Gedmatch (EuropGenes JTest I score 5.33% Ashkenazi, about 5% West Asian and 5% East Med).
On FTDNA's old PF I had no ashkenazi (the 4 populations were West Europe 65%, South Asia, East Asian and West Africa). Now FTDNA myOrigins gives me 7% jewish diaspora (ashkenazi) and 4% middle east added to the previous 4 populations. It reduced my 65% western euro to 55% (-10% = to 7+3+/-1% error that are now in JD and ME).
So anyway, it seems to agree with Gedmatch. By the way most of my euro is north mediteranean meaning south of europe and then british isles).
Now comes the wierd part. 23Andme has the same 4 population slip (euro, africa, south asia, east asia), but 40% of my euro is unspecified. I have no ashkenazi matches whatsoever on gedmatch, FTDNA or 23Andme. A few J2, J1 YDNA on 23andme and 1 N1b mtDNA.
If it was noise (most probable reason), why would gedmatch and myOrigins pick it up (by the way i forgot to mention that Dr Doug MacDonald also shows some middle east on my chromosome painting) and yest i would have no ashkenazi match? On gedmatch i have a few ashkenazi matches at 6.8 cM, but nothing above 7cM?
Well I do not if i should consider this to be true of just some noise?
by the way i have a few sephardic ancestors from 350 years ago!
thank you

ADW_1981
05-22-2014, 12:59 PM
Thank goodness 23andme does chromosome painting, and I can see they match me on the same spot of my chromosome 7 that is painted Ashkenazi. My Jewish matches I can see that same chromosome for them is overwhelming Ashkenazi on top and bottom, but especially in the spot they match me.

I may have asked you this before, but is your Y lineage part of the Ashkenazi Jewish cluster? Or the lone NW European one (perhaps the result of a wandering Khazar way back when?) It seems I'm related to (several) of the fellows part of the same NW European cluster and the "public" guy appears 100% European in myOrigins, despite being Q1b1a. Interesting nonetheless.

Silesian
05-22-2014, 01:58 PM
My Ashkenazi admixture is only 4.7% in standard or 6.1% in speculative. What seems like Jewish surnames dominate my DNA relatives, just as AJL was explaining. Schwartz is a common German surname by the way.


surname closest relative count enrichment
Levin 4th Cousin 8 103
Cohen Distant Cousin 10 76
Schwartz 4th Cousin 8 70
Katz 4th Cousin 5 52
Goldstein Distant Cousin 5 52
Levine 4th Cousin 5 50
Goldberg 4th Cousin 5 48
Brown 4th Cousin 6 5
Davis 4th Cousin 5 4
Smith 4th Cousin 5 0

MiTuCents has a good thread with 2000+ posts on Ashkenazi heritage and admixture. Have you posted your results so he can assess your Ashkenazi heritage ?

MiTuCents
MiTuCents Do You Have Ashkenazi Jewish Admixture?

Find & Post Your Values & Determine your Level of Ashkenazi Admixture here:

FIND & POST YOUR VALUES:

Example:

Ancestry Composition (Standard Mode): 49.2%
Declared Ashkenazi Jewish (5cM, 1+, US Incl): 29.4% - 38.8%
Total "DNA Relatives" Matches: 1122
K1a1b1a Matches in "DNA Relatives": 191
K1a1b1a Percent (191/1122): 16.0%
Chip Version: v3

Actual Ashkenazi Admixture (If Known): 1/2

Update percentages:

*** RECENTLY UPDATED RANGES (20Mar2014) ***

1/32 AJ (1 AJ GGG Grandparent):
AC: 1.4% - 4.9%
COA: 0.6% - 4.0%
K1a1b1a: 3.9% - 12.1%

Táltos
05-22-2014, 02:11 PM
I may have asked you this before, but is your Y lineage part of the Ashkenazi Jewish cluster? Or the lone NW European one (perhaps the result of a wandering Khazar way back when?) It seems I'm related to (several) of the fellows part of the same NW European cluster and the "public" guy appears 100% European in myOrigins, despite being Q1b1a. Interesting nonetheless.
Yes so far it would appear that my father line is part of the Ashkenazi Jewish cluster. My brother is DYF395S1=15-19, which is the AJ1 cluster, and he is positive for SNP Q-L245. Right now I am awaiting his result at YSEQ, for a private SNP that was found in a AJ1 sample.

I just posted a link to the updated Q1b paper, see here http://www.anthrogenica.com/showthread.php?1848-Phylogenetic-Structure-of-Q-M378-Subclade-Based-On-Full-Y-Chromosome-Sequencing Dr. Gurianov and his colleagues are really making a lot of progress in understanding Q1b.

AJL
05-22-2014, 03:37 PM
thanks

so the X chr which in 23andme has a small Ashenzaki marker which matches Ch6 in 23andme a a small Ashenzaki ( same numbers ) and matches exactly my DougM 's X chr as being South-Asian , can mean basically nothing in terms of finding any links to people?

Can you link anything about chr 6 being the immune sytem?

Yes, it seems very hard from what you have said to conclude anything concrete. We must remember that there are some "strange" (for European) yDNA in the Alps, such as H and L, not found in most of the rest of Europe. This might account for some ancient South Asian–like link.

http://genome.wellcome.ac.uk/doc_WTD020716.html

AJL
05-22-2014, 03:44 PM
Now comes the wierd part. 23Andme has the same 4 population slip (euro, africa, south asia, east asia), but 40% of my euro is unspecified. I have no ashkenazi matches whatsoever on gedmatch, FTDNA or 23Andme. A few J2, J1 YDNA on 23andme and 1 N1b mtDNA.
If it was noise (most probable reason), why would gedmatch and myOrigins pick it up (by the way i forgot to mention that Dr Doug MacDonald also shows some middle east on my chromosome painting) and yest i would have no ashkenazi match? On gedmatch i have a few ashkenazi matches at 6.8 cM, but nothing above 7cM?
Well I do not if i should consider this to be true of just some noise?
by the way i have a few sephardic ancestors from 350 years ago!
thank you

Hello,

Have you run Ancestry Composition on Speculative view as well as the standard one, and if so were the results much different?

Given what you say it is possible you have some Near Eastern or nearby but non-Ashkenazi ancestry. This might be Sephardi, Mizrahi, Samaritan, Arab, Druze, Assyrian, Bedouin, or even something like Cypriot or North African, or some mix thereof. What does your Country of Ancestries show? Is it possible your Sephardi is a greater proportion of your ancestry than 350 years ago?

esev
05-22-2014, 04:46 PM
Hello AJL and thak you for the answer.
AC standard:
60.3% euro
0.2 british
10.8 unspecified northern

2.8 iberian
6.2 nonspecific southern
40.3 % unspecified euro

the reste is african sub sahara, south asian and east asian/native american.

on speculative
my iberian shouts up to 10% and italian at 3.

AC in speculative mode just increases a lot my south europe (iberian and italian) and decreases a bit my unspecified. but no it has no ashkenazi. Most calculators on Gedmatch are puzzled by my data. They all seem to see around 10% of caucasus region (armenia etc) and 2.5% arabian, bedouin. So they get mixed up in the calculation and all oracles always end up giving me around 5 to 8% of either Ashkenazi, Druze, Yemenite jews, morocan (because i score 12% west african probably). So yes it does appear that i have some old left over genes from the middle east but a bit more from the caucasus. Maybe that's were they all get confused! I followed MiTuCents instructions on 23andme and my results would not indicate any recent ashkenazi ancestry:
AC 0%
COA 01-1.1% (self declared ashkenazi people from 1 GP to 4 GPs)
mtdna often ashkenazi:
K1a1b1a 0 out of 330 matches
K1a1b1 1 out of 330 matches
k1a1b2a 1 out of 330 with a J2 YDNA???
N1b 1 out of 330

YDNA:
j1 3 out of 330
j2 9 out of 330
e1b1b 5 out of 330

Still trying to figure out why JTest has me at 5% ashkenazi and myOrigins 7% ashkenazi!

regrads

AJL
05-22-2014, 05:06 PM
Thanks for answering. It's certainly possible as you say that you have some non-Jewish but other West Asian or Caucasus-area ancestry, which would fit well with J2, which is combining with your traces of Sephardi and African to make you seem more Near Eastern than you actually are. The more complex/admixed a person is, the harder it is for the algorithms to deal with it.

esev
05-22-2014, 05:44 PM
Thank you AJL. I have considered that possibility. Yet there might be another one. You see 23andme looks at the last 500 years. Myorigins goes into deep ancestry example 1 to 2000 years for ashkenazi (seen on there forums), same as gedmatch. They both pretend to look at deep ancestry and strange enough they give me close ashkenazi results 5 to7 %.
23andme 500 years span confirms what i know from my family tree, we havent had ashkenazi enter the tree for the last 350 years at least. As i said only a few sephardics but again 350 years ago only. So that maybe why 23andme does not pick up anything while myorigins and gedmatch can still see a bit of it. It could well be in that 40% unspecified euro which according to 23andme means belongs to genes that are shared by more than one specific population hence unspecific.

Regards

AJL
05-22-2014, 06:09 PM
Yes, something like that is certainly possible, especially if you happened to pick up a fairly large chunk of Sephardi ancestry. It may also be worth considering the role of the Phoenicians in Iberia.

vettor
05-22-2014, 06:48 PM
Yes, it seems very hard from what you have said to conclude anything concrete. We must remember that there are some "strange" (for European) yDNA in the Alps, such as H and L, not found in most of the rest of Europe. This might account for some ancient South Asian–like link.

http://genome.wellcome.ac.uk/doc_WTD020716.html

thank you

And yes, H and L are in the alps along with my T marker , who was once joined in haplogroup LT ............

If the DougM term of south-asia is correct in my X chr and sits ( as well as Chr 6 ) in exactly the same spot as 23andme who names it ashkenaki , then most likely it has to be purely a maternal line from a maternal side of that line

BTW , 23andme lowered my ashkenaki yesterday from 0.7% to 0.4% ( standard) and increased my Irish portion 0.3%...............are they watching!

vettor
05-22-2014, 07:07 PM
maybe I have some small Ashkenaki............i recall this below
T ydna markers in northeast Italy and western Austria

http://i103.photobucket.com/albums/m153/vicpret/mySNP_zpsd9e61f4b.jpg (http://s103.photobucket.com/user/vicpret/media/mySNP_zpsd9e61f4b.jpg.html)

I am the far right one .................and note the other 5 types of less than 1%

esev
06-02-2014, 07:19 AM
Dear all, following the discussion we had las time on Ancestry Composition I found the following on the net:
here is the link first
http://dienekes.blogspot.fr/2012/12/23andme-ancestry-composition.html

If this is true, and i am not in any position to ague here since i am not competent in that field, this could raise some serious question as to the accuracy of AC!
I quote:

"Most of the reference dataset comes from 23andMe members just like you. When someone tells us that they have four grandparents all born in the same country, and the country isn't a colonial nation like the US, Canada or Australia, they become candidates for inclusion in the reference dataset. We filter out all but one of any set of closely-related people, since they can distort the results. And we remove "outliers," people whose genetic ancestry doesn't seem to match up with their survey answers.
23andMe takes a "birthplace of grandparents" approach rather than an "ethnic origin" approach."

This definately cannot be accurate. For instance, when i answered this survey, i had no choice but to answer France for my 4 grand parents because the place i come from was not in the list, that place being the french island of Reunion in the Indian ocean, where we are french by nationality but not 100% genetically (we are euro, indian, african and asian).
So for my case it makes the survey inaccurate!

Regards
Eric

AJL
06-02-2014, 03:54 PM
Eric:

Relative Finder is not the same as Ancestry Composition. RF looks at longer segments -- though in many cases, still not long enough to be of genealogical use. I have had better luck with Family Finder, but even that will be plagued by a problem of "pseudosegments."

Ann Turner analyzed a whole whack of phased data and found that false segments can be quite a bit larger that generally thought:

http://www.jogg.info/72/files/Turner.htm

AC's segments are smaller than either, which of course also raises questions of validity. If a substantial number of ~7.5 cM blocks are spurious therefore untraceable, what does that say about the shorter 100-SNP (albeit phased) blocks used for AC?

Still, if we are talking about general regions instead of specific populations, AC remains reasonably accurate, if not always particularly helpful (a lot of people have an awful lot of "Nonspecific").

Ann Turner
06-03-2014, 11:13 AM
Eric, your results almost certainly would not be included as part of the reference dataset for Ancestry Composition. The first step in assembling the reference populations is to remove "outliers" -- results that don't cluster together.

AJL, the 100-SNP blocks are a "sliding window." The Ancestry Composition algorithm will be looking for consistency in adjacent blocks.

warwick
06-03-2014, 05:15 PM
A large number of customers of Ashkenazi descent have bought 23andme, both for health and ancestry reasons.

Here's why Relative Finder is useless for Ashkenazi ancestry.

In one entirely Ashkenazi person's profile, these surnames are most common among matches:

1. Cohen
2. Katz (variant of Cohen)
3. Levine
4. Goldberg
6. Friedman
7. Stern
8. Greenberg
9. Segal (variant of Levine)
10. Horowitz (variant of Levine, Segal)
11. Lifshitz
12. Finkelstein
13. Shapiro (variant of Cohen, Katz)
14. Schwartz
15. Epstein

Compare these to the most common Ashkenazi surnames of arrivals at Ellis Island in the late 19th century/early 20th, courtesy Yannay Spitzer (http://yannayspitzer.net/2012/07/24/most-common-jewish-names/):

1. Levin/Lewin (variant of 3 on list)
2. Kaplan (variant of 1,2,13 on list)
3. Goldberg (4 on list)
4. Katz (2 on list, also variant of Kaplan etc.)
6. Goldstein
7. Cohen (1 on list)
8. Schapiro (13 on list, often also a variant of Cohen etc.)
9. Epstein (15 on list)
10. Rosenberg
11. Friedman (6 on list)
12. Weinstein
13. Schneider
14. Feldman
15. Kagan (variant of 1,2,13 on list)

Ah well, the lists are almost the same. What does this mean?

In brief, almost all links are specific to a gene pool rather than a family. There is no way to discern actual ancestral ties via shared surnames for anything closer than about a second or third cousin, as one would expect since most of these surnames were independently adopted about 1770. In the same person's kit, there are 149 predicted second-to-third cousins among 3,065 DNA relatives, none of whom is a 2nd/3rd cousin by paper trail.

Let's also take a closer look at some of these surnames and how they correlate to haplogroups:

Horowitz: J1, Q1b, R1a1a, T
Levin, Levine: E1b1b1c1a, G2a4, G2c, J1, J1e, J2, J2a1b1, R1a1a, R2
Segal: R1a1a, R1b1b2
Cohen / Kaplan / Katz / Kagan: E1b1b1c1, E1b1b1c1a, G2a4, G2c, J1, J1e, J2, J2b2

Need I say more? :\

My view is that relative finder type tools have very limited utility for Ashkenazi ancestry, primarily because most Ashkenazi Jews have a high probability of sharing haploblocks even with people 500 years in the past, since the founding population was only about 1,000 people in the year 1000. This tool and others like it simply show that one descends from that founding population. Another problem is that most of the paper records have been destroyed or were not maintained, so it is difficult to pursue the research before 1800 or so.

This isn't a limitation of RF alone. It applies to every tool that uses autosomal DNA for Jewish ancestry. Given the small size of the founding population and the high likelihood of sharing, most of the matches are useless.

I don't regard this as a criticism of 23andme. However, it might be wise to make a stronger caveat about the meaning of those 1,000s of matches that the typical Ashkenazi finds at 23andme or elsewhere. They are not likely to be genealogically useful.

warwick
06-03-2014, 05:22 PM
Essentially, if you're trying to connect genetic results with actual paper trail documentation, I think the problems for Jewish genetic genealogy are enduring and persistent. The paper records are poor prior to 1800 and many records have been destroyed since then. Also, many surnames were not adopted until the time of Napoleon, which also makes research quite problematic.

Until the 19th Century most Ashkenazi Jews had no family names. Instead, they were named after their parents
http://www.bje.org.au/learning/jewishself/family/familynames.html

The problem is a tough one.

seferhabahir
06-03-2014, 11:34 PM
Essentially, if you're trying to connect genetic results with actual paper trail documentation, I think the problems for Jewish genetic genealogy are enduring and persistent. The paper records are poor prior to 1800 and many records have been destroyed since then. Also, many surnames were not adopted until the time of Napoleon, which also makes research quite problematic...The problem is a tough one.

If this is an interest area for some people, consider attending the upcoming 34th IAJGS Iternational Conference on Jewish Genealogy in Salt Lake City... (not very deep on genetics yet, but a lot of paper trail talks). Next year in 2015 it is in Jerusalem, and the year after in 2016 it is in Seattle.

http://conference.iajgs.org/2014/about_conference.cfm

Goldschlager
06-04-2014, 12:03 AM
Completely agree as to the usefulness (or lack of) of autosomal test -- my list of Family Finder matches has 2,634 entries out of which 340 are for suggested 4th cousins and closer (and this supposedly after FTDNA has applied their algorithm for endogamous populations). I have a paper trail to early 1800 on all my lines. My hope in advancing this is entirely based on the Y DNA test. There are lucky people, Jewish or Gentile but with Jewish ancestry , whose paternal ancestor lived in the earlier periods in places like Germany or some Austrian areas where records have been kept. Several of my close (GD 1 and GD 2) matches belong to this. The challenge is to find somehow the "connection" point.

AJL
06-04-2014, 12:35 AM
^ One of the more interesting things I have discovered from my non-Ashkenazi matches with Ashkenazi ancestry is that I match both a Dane and a Finn who have ancestors named "Pfeiffer." This makes it quite likely that I have an ancestor by that name.

But testing multiple Y lines to high resolution is probably most beneficial.