Ancestry Autosomal Results

[Mlawton]

Okay, I hope this doesn't turn into a fight about Ancestry, but I could use some advice and expertise.

I have a 4th cousin who recently submitted an autosomal kit at Ancestry. She is new to dna and we had spoken awhile back about the different companies offering this test. Her results are back at TWO weeks. She is very leery about the accuracy and so am I. Our common line, Traylor and Sparks is nowhere to be found among her matches. Not one Traylor and not one Sparks. In addition, she is not a match to me, nor do I have her on my list of matches, which was updated yesterday with 48 low confidence distant cousins. However, I have two confirmed Traylor matches there, both of which have the exact same relationship to her. One of them is also a 4th cousin, all three of our gg grandparents were siblings. In addition, I have a number of Sparks distant cousins, which she should also, as our Traylors married Sparks. The Sparks line especially should be represented in her distant cousin matches, particularly because they are quite an extensive documented family and I have a number of them at Ancestry as well as here.

While I understand that we might not inherit the same amount of dna with each generation, I believe that at the very least I and the other 4th cousin should show up on her match list, even if the more distant cousins that are on my list do not. And certainly at least one or two of the numerous Sparks should at least show as distant cousins. Going through every single match, she is unable to find a single person for either her paternal or maternal side.

Two weeks is an unbelievable turnaround. My Ancestry test was just a bit over 4 weeks, with my Ftdna autosomal all back right at about 5 weeks give or take a few. How could Ancestry have results so quickly and why does she not have any matches to our comment documented lines? Ideas?

[MMaddi]

The main thing to keep in mind with autosomal matching to detect cousin relationships is that as the cousin relationship is more distant, the less chance there is that the cousins will share a large enough segment to detect. That's the case for autosomal tests with any company - Family Finder, 23andMe or AncestryDNA.

Take a look at the FTDNA's FAQ at http://www.familytreedna.com/faq/answers.aspx?id=17#628 which gives you the probabilities involved. Up through 3rd cousins, an autosomal test will detect the relationship in 90% or more of the cases. When you reach the 4th cousin level, only 50% of 4th cousins will share large enough segments of DNA for the test to catch and tag as indicating a cousin relationship.

What you've described above involves all 4th cousin relationships. There are different siblings involved at the gg-grandparent level. So, what has likely happened is that in the case of you and the 4th cousin you don't match, the siblings did not both inherit the same segments and/or in the generations since then recombination has cut up or eliminated any segments the siblings at the gg-grandparent level shared. In the case of the 4th cousin whom AncestryDNA has tagged as your cousin, at least one shared segment from the sibling gg-grandparents has been preserved over the generations and is still large enough in you and that 4th cousin for the test to detect.

It's all based on a roll of the dice in each generation. And at the 4th cousin level, the odds are 50/50 that shared segments will be preserved.

[Mlawton]

Thanks! I do understand the recombinant part and that the amount of dna passed on to the next generation diminishes and/or mutates, but Ancestry gives me a 96% and 95% confidence levels for my fourth cousin matches, as well as the one who shares a 3 X gr grandparent. I guess what I'm asking is why a single Traylor or Sparks does not show in her matches at all even distant. Surely one of those should pop up even if say, a 5th or more distant Sparks with a low confidence level. I have a number of these Sparks distant cousins as they are as common as fleas. (maybe I shouldnt use that analogy). But there's a heck of alot of Sparks descendants out there, and I have a number of distant matches of them here and at Ancestry. Not even a one of them matched to her?

I guess what I'm saying is that I know there are numerous Traylors and Sparks in the database there, and she doesn't have one match to any of them. Just doesn't sound right. She has told me she is positive she is the daughter of both of her parents. I guess one possibility is that one of her paternal ancestors between our GG granfathers, is not from the Traylor Sparks connection. Does that make sense. I wonder if Ancestry would rerun the results... Two weeks is pretty quick in my mind....

[djknox]

MLawton: I can't offer any expertise, but some context might be helpful. Put another way, for many years genealogists had only paper trails to research their pedigrees... which, for current generations, tend to fade quickly at the 4th to 6th cousin level (early 19th century back into the mid 1700s). So along comes autosomal dna testing, which can offer matches at the 5th cousin level 10% of the time, and less than 2% of the time at the 6th cousin level. At least this is what the algorithms are capable of - I suspect advanced and customized analysis may improve those odds... but Ancestry is yet to provide its clients real data for download.

One can thus conclude these tests are not much of a tool for advancing one's pedigree in the very zone paper trails begin to fail. What autosomal dna testing can (only at times depending on regional specifics of your ancestors) do, is confirm paper-trail findings in the 2nd to 4th cousin range. In your case however it would appear not. (its also effective for recent adoption and false paternity... of which I seem to have none)

Note that Ancestry is claiming some "phasing" alogoritms that should offer better results than ftdna... but thus far I for one haven't seen results that would indicate that. I must have a 1000 matches at Ancestry and have not linked one to my tree. I have also had one distant cousin test and he failed to show up on my matches. FTDNA is also rumoured to be introducing phasing into their autosomal product... but its anyone's guess when it will be released and whether it will improve matching success significantly.

Against my instincts (due to the high costs involved), i'm investing in the testing of maternal and paternal uncles and aunts, and siblings, to see if some sort of "triangularization" of my dna ancestry can help add better "resolution" to matching at the 4th to 6th cousin level. T Peterson on this forum advocates using tests from 1st cousins too - apparently he's had some success with that. There are also some who advocate admixture (ie population finder) analysis for helping one's research... but IMHO good luck with that too - its like painting portraits with a roller brush.

Anyway, you're not alone in your disappointing results. Maybe when Ancestry makes the raw data available there might be some expertise analysis that sheds more light on your situation... but as MMaddi states, due to recombination probabilities, you still might be S out of luck.

[Mlawton]

No, I personally have had stellar results with my kits here and even at Ancestry where a match proved my great grandmother through a different line (her half sibling's line) that was also proved here by testing a 3rd cousin. That was a brick wall that I had for almost 15 years. I have proved paper trails far beyond the 4th cousin relationships I mention above, notably in my Sparks relatives...the Sparks line being a very easy line to document with paper trails.

I have had nothing but success with my autosomal testing and ydna and am a huge fan of this type of genealogical inquiry, although I am certainly no expert and have a rudimentary understanding compared to most here. In this case I think that the results for this 4th cousin are defective and incorrect. We should have at least one match in common somewhere with the amount of both Traylors and Sparks in the database there. I guess there could be a NPE somewhere, but I think she should contact Ancestry and ask for a redo. Maybe I could convince her to test here...if we match here, there is definitely a problem with her Ancestry results.

[travers]

It doesn't matter if you have well documented lines for your Sparks family if their descendants haven't tested also. If most of the ones that are related through those lines that might have tested are distantly related (5th or further back) than the odds are most if not all will not match. You didn't match your 4th cousin, well there was only a 50% chance you would. It doesn't mean the test is defective that's just how it is.

The best use of the autosomal tests are for confirming 3rd cousins or closer and recruiting specific people within these confines to confirm relationships. Ancestry is not the best for that strategy as they are not giving you segment matching info. I have tested with ancestry too though just to see what random matches I will receive. Most of the matches these types of tests are going to randomly pick up are very distant cousins, that's because obviously you have tons more 8th, 9th, 10th cousins than you will have 3rd and 4th or closer cousins. You have so many ancestors (thousands) that far back in your tree it is almost impossible to figure out how you will be related to most.

It's possible for two people to match here and not there or vice versa as all these companies have different matching algorithms, but they will all pick up the close relatives. Getting your results back quickly doesn't have anything to with anything and I highly doubt your 4th cousin's results are "defective". I doubt ancestry will redo the testing just because you ask. If she's so unsure just have her test a close family member. A parent or a grandparent or older relative tested on the side of the family she wants to learn more about would be helpful anyways.

[Mlawton]

Okay. I won't belabor your points. I do understand what you are saying and appreciate it. But I guess that when you say 50% probablility of matching at
4th cousin, and there are multiple 4th cousins, this 50% probablility is not cumulative but is per event, right? Because that is what I am having a hard time getting. She doesn't match me, her 4th cousin, and she doesn't match our other 4th cousin whom I match, and she doesn't match another cousin 4th 1X removed. In addition, she doesn't match any Sparks descendants that are in Ancestry's database, of which there are many, albeit distant cousins. And this is because the probability or % of chance of finding a relationship at these distances is calculated singly and non-cumulatively. So the more Sparks and Traylors that are in the database does NOT statisticaly increase the odds of finding at least one match. for example if I have 10 Sparks distant cousin matches, and her relationship to those 10 are the same as me, her odds of the program matching her to at least one of those 10 is 50% or less each occurence? Her odds of making a match do not increase with increasing numbers of cousins to the database? Be it 1 or 100 Sparks, her chance of finding a match is the same if only one Sparks is in it or 100 Sparks. Is this correct? I dropped out of Statistics in college because it was too HARD! thanks for your patience.

[djknox]

A quick explanation concerning your comments about probability in this specfic example. If one has many 4th cousins of a Sparks ancestor, you are correct that accumulatively the probability is that many should match each another... and actually, provided one matches with one of them would indicate it is likey they would match with many of the others. HOWEVER, in the event that one individual did not retain sufficient dna from the Sparks line ( ie the specific 4th cousin you are discussing has lost most of his/her Sparks dna through recombination), then that person will NOT match the other 4th cousins decendent from the same Sparks ancestor... even if there are 100s of them.

Without getting too boring wrt probability analysis, if 50% of Sparks 4th cousins typically match, and there is 100 4th cousins, through recombination distribution there is a near certain probability that at least 1 of the 100 no longer retains Sparks dna sufficient to match the other 99. It just happens that this 4th cousin of yours belongs to that subset.

To try to explain this with a rough analogy, if you take 128 people and ask them to each flip a coin 6 times (each time is like a generation), on average 50% will be heads and 50% will be tails. However, there's a very good chance that 1 of them would have flipped 6 heads in a row, and 1 of them would have flipped 6 tails in a row, while the remaining 126 would have some combination of heads and tails. In this case, 126 people have some similarity to each other... but the 127th and 128th persons have no similarity with each other. Recombination is a little different than this analogy... but if you define Sparks ancestry as having a heads, then there is 1 person in 128 that have zero dna common with the Sparks. If the resolution of matching is such that one needs at least 2 out of 6 heads to show up as a Sparks, then there's a whole bunch of cousins within the 128 group who will not match to Sparks. In your case, it is reasonably possible that your 4th cousin is one of several who flipped heads zero or 1 time only.

I hope this helps.

[tomcat]

...
I have a number of Sparks distant cousins, which she should also, as our Traylors married Sparks.
...

I doubt Ancestry will send a new kit but suppose they might check their phasing and matching routines in this case of intermarried families as they might learn something useful.

[tomcat]

When Ancestry releases raw data next year it can be ported to GedMatch to see how it performs under a different matching routine.

[tomcat]

Here are my ancestry estimates from AncestryDNA:

• British Isles 19%
• Persian/Turkish/ Caucasus 13%
• Native North American 13%
• Southern European 13%
• Middle Eastern 8%
• Central Asian 8%
• Eastern European 21%
• Uncertain 5%

I am paternally, ethnically Jewish, with two Ukrainian GP's and most of my AncestryDNA matches have European Jewish as a primary ethnicity. No surprise there. It is somewhat surprising AncestryDNA did not so classify me, even with the hint of my profile.

So, my results are those of a half European Jewish person not specifically identified (by AncestryDNA) as European Jewish.

[NYMark]

I have noticed quite a few matches on Ancestry with distinctly Jewish surnames in their trees who are not designated as Jewish by Ancestry. Some of these are apparently more than 50%. On the other hand Ancestry seems to have gotten my admixture just about right.

Stopped clocks and all that. . .

Here are my ancestry estimates from AncestryDNA:

• British Isles 19%
• Persian/Turkish/ Caucasus 13%
• Native North American 13%
• Southern European 13%
• Middle Eastern 8%
• Central Asian 8%
• Eastern European 21%
• Uncertain 5%

I am paternally, ethnically Jewish, with two Ukrainian GP's and most of my AncestryDNA matches have European Jewish as a primary ethnicity. No surprise there. It is somewhat surprising AncestryDNA did not so classify me, even with the hint of my profile.

So, my results are those of a half European Jewish person not specifically identified (by AncestryDNA) as European Jewish.

[tomcat]

I wonder if the difference is between those who elected to be part of the ancestry project and those that did not, wherein those not electing participation got the summary reading and those electing participation got the the full range?

[EdwardRHill]

I hear Jessie Ventura is considering doing a show about AncestryDNA. I heard that its has something to do with space ship and little green men and area 57. DOH!!!

[Yaffa]

Okay. I won't belabor your points. I do understand what you are saying and appreciate it. But I guess that when you say 50% probablility of matching at
4th cousin, and there are multiple 4th cousins, this 50% probablility is not cumulative but is per event, right? Because that is what I am having a hard time getting. She doesn't match me, her 4th cousin, and she doesn't match our other 4th cousin whom I match, and she doesn't match another cousin 4th 1X removed. In addition, she doesn't match any Sparks descendants that are in Ancestry's database, of which there are many, albeit distant cousins. And this is because the probability or % of chance of finding a relationship at these distances is calculated singly and non-cumulatively. So the more Sparks and Traylors that are in the database does NOT statisticaly increase the odds of finding at least one match. for example if I have 10 Sparks distant cousin matches, and her relationship to those 10 are the same as me, her odds of the program matching her to at least one of those 10 is 50% or less each occurence? Her odds of making a match do not increase with increasing numbers of cousins to the database? Be it 1 or 100 Sparks, her chance of finding a match is the same if only one Sparks is in it or 100 Sparks. Is this correct? I dropped out of Statistics in college because it was too HARD! thanks for your patience.

I hope this helps you some. my 1/2 sister and I tested FF. We share 1800-1900 cm of DNA. We share same father different mother. My 1/2 sister and I share some same DNA we got from our father and some different. She has matches from my father that I don't and visa versa.


Also I have family tested on my mother's side. I have a 4th double cousin line who does not match our family on FF. I had FTDNA double check and we do share a very small amount of DNA with this double 4th cousin but it is under what FTDNA considers a match and why we are not showing up as a match on FF


also in general the Sparks family took over the south. Especially western NC. not all those with the Sparks surname may be related. my ancestors got tons of Sparks neighbors.