measurement error in DNA testing?

[georgian1950]

I'm just getting started in learning about DNA as it applies to genealogy. I have one simple question which I can't seem to find the answer to and that is what sort of measurement error can you get with DNA analysis. For instance, if you retook the 37 marker Y chromosome test with a new sample, would you be say 99.99 percent confident that you would get the exact same results, or would it be somewhat less? I know something like contamination of samples could affect results, but the error would be gross enough that it would be noticed. Could you routinely get a small error on a marker or two? I know that would not be enough to affect your overall classification, but it could affect your hunt for exact and close matches.

I appreciate any insight to this question.

[Brunetmj]

The only way to know the answer to this would be to submit multiple kits of the same kit and check results . Who has the money for that.
I can say a national geographic test yielded the same 12 marker values as FTDNA. However a person could point out they are affiliated.
23andme results , which gives deep clade on the initial test matches my deep clade here. However 23andme does not yield STR values.
I suppose those who had a brother or first cousin tested could offer details but not sure why they would Y test.
I guess how accurate they are could only be inferred but not proven.

[teyoung]

For instance, if you retook the 37 marker Y chromosome test with a new sample, would you be say 99.99 percent confident that you would get the exact same results, or would it be somewhat less?

I Y tested at both FTDNA and SMGF and received identical results for the common markers. My SNP result from Deep Clade at FTDNA was verified by 23andMe.

I would say the accuracy is very, very good.

There was a mixup at 23andMe, but it was a result of a sample tray being reversed, or something like that. The errors were obvious, like females getting a Y haplogroup.

[thormalen]

I was e mailing with a 67 marker match. The person had supplied her brother and cousins Y DNA. She said that FTDNA must have made a mistake as there were no Grahams in her family as far back as she knew and all their matches were with Grahams. I replied, but both your brother and your cousin are matches with me. Surely the same mistake was not made with both....... The mistake is that people get so overwhelmed by the trueness and importance of their family name that sometimes it is hard for them to believe that somewhere down the line a male child was sired out of wedlock, or an unrecorded adoption took place or the last name was changed for one of many reasons and the name they are so proud of does not match the blood line. Welcome to the real world. That is part of the fun of doing DNA research.

[similligan]

I originally had my father tested at DNA Heritage and then retested at FTDNA. This was before DNAH sold out to FTDNA, so it wasn't just a conversion. All markers the two companies tested in common came out the same.

I also have a few project members whose results didn't match the lineage they expected. Since there would always be some doubt in their minds, one of the suggestions I give is for them to order another 12 marker test. One ordered under an alias, another asked for a retest. In both cases, the results were as originally reported.

Susan

[josh w.]

I'm just getting started in learning about DNA as it applies to genealogy. I have one simple question which I can't seem to find the answer to and that is what sort of measurement error can you get with DNA analysis. For instance, if you retook the 37 marker Y chromosome test with a new sample, would you be say 99.99 percent confident that you would get the exact same results, or would it be somewhat less? I know something like contamination of samples could affect results, but the error would be gross enough that it would be noticed. Could you routinely get a small error on a marker or two? I know that would not be enough to affect your overall classification, but it could affect your hunt for exact and close matches.

I appreciate any insight to this question.

Is there a reason why you are asking the question, e.g. do you have doubts about some results. Commercial testing companies are unlikely to publish results on their own reliability, but my guess is that there have been internal studies. The 99.99% confidence interval would seem to be extremely high for most medical measurement.

[georgian1950]

Is there a reason why you are asking the question, e.g. do you have doubts about some results. Commercial testing companies are unlikely to publish results on their own reliability, but my guess is that there have been internal studies. The 99.99% confidence interval would seem to be extremely high for most medical measurement.

I took the test to get past a brickwall with my GGG-grandfather in trying to attach to one of the known lines of my surname. I got past the brickwall all right, but in the process I found that someone had corrupted the family line. I'm not questioning that - there's no doubt. Right now, I'm just trying to establish which surname I need to latch onto next. The evidence suggests an answer, but the deed would be 300 years ago and seven generations on my side and probably more than that on the other side. On the 37-marker test, the genetic distance is not quite close enough to establish the link, but certainly is closer than you would expect at random. A mistake of a few points in the genetic distance would help the case.

I agree that the testing companies would have their own internal studies. If the confidence was 99.99%, I think they would let us know.

[georgian1950]

I may have been premature in starting this thread. A few days ago I asked FTDNA this question. I finally got a reply back this evening. The company's retesting policy is to give a retest at a discounted price. If the results come back differently, it will refund the testing fee. In 12+ years of business, FTDNA has not had to issue a refund.

This makes me a bit curious about the science behind the testing when complex results can be produced essentially error free.

[Ann Turner]

The only errors I've heard about have been due to sample mix-ups. The DNA measurements themselves are really very robust.

For some types of measurement, we do expect slightly different results when the test is repeated. For example, suppose you are standing on the 5th rung of a ladder and I measure your distance to the ground ten times in a row to the nearest 1/16th of an inch with a measuring tape. You should be suspicious if I give you exactly the same number every time. But if you asked me which rung you are on, you should expect me to give you the same number every time.

Measuring the number of repeats is more akin to counting the rungs on the ladder.

[jbarry6899]

Errors in Testing

The error rate for identifying repeats in markers should be very low, but the calculation of time to most recent common ancestor (TMRCA) is dependent on assumptions regarding mutation rates, and can vary depending on the precise figures used. FTDNA's calculator is said to be very sophisticated. However, you may want to explore several different calculators to see how the results differ. Just Google TMRCA calculator.

Also, you will want to decide what level of probability to use to underpin your investigations--50%? 75? 90? My own inclination is to try to research any match that shows up within genealogical time frame (after surnames became common) that has a probability of 67% or more.

[josh w.]

I took the test to get past a brickwall with my GGG-grandfather in trying to attach to one of the known lines of my surname. I got past the brickwall all right, but in the process I found that someone had corrupted the family line. I'm not questioning that - there's no doubt. Right now, I'm just trying to establish which surname I need to latch onto next. The evidence suggests an answer, but the deed would be 300 years ago and seven generations on my side and probably more than that on the other side. On the 37-marker test, the genetic distance is not quite close enough to establish the link, but certainly is closer than you would expect at random. A mistake of a few points in the genetic distance would help the case.

I agree that the testing companies would have their own internal studies. If the confidence was 99.99%, I think they would let us know.

The dna information is 99.99% more reliable than any family genealogical information. Be aware that genetic distance measures point to a wide range of dates (in terms of number of generations) rather than a specific point in time. I am not sure what you mean when you say the line was 'corrupted'. Do you mean a non paternal event---this might have strengthened the family line.

[sjadelson]

FWIW, I originally tested with Ancestry's Y46, and my results at FTDNA matched exactly.

[georgian1950]

The dna information is 99.99% more reliable than any family genealogical information. Be aware that genetic distance measures point to a wide range of dates (in terms of number of generations) rather than a specific point in time. I am not sure what you mean when you say the line was 'corrupted'. Do you mean a non paternal event---this might have strengthened the family line.

Yes, a NPE. I guess I need to use the right terminology. It looks like a wife of someone in my family line let a man other than her husband into her bedroom (or barn?). I think I have a good guess as to who it might be, but the genetic distance isn't close enough to convince most students of genealogy. But we are talking 300 years back to the event and most of the mutations are in the more rapidly changing markers. I don't follow what you mean by a how a NPE might strengthened the family line.

Anyhow, I appreciate all of the responses and I think this thread has pretty well accomplished what was intended for it. I particularly find Ann Turner's analogy with the laddar very useful.

I know all of us start out as novices and have to learn a lot. I don't want to bore forum members with my own problem, but the way it appears, there were a bunch of NPE's happening back in that late Colonial - early post-Revolutionary War period. My problem might be a good case study to present, and might help people from several family lines get past their brick walls. Does anyone think I should start a new thread with the details of my situation?

Thanks again for all of the help.

[slwr]

Yes, a NPE. I guess I need to use the right terminology. It looks like a wife of someone in my family line let a man other than her husband into her bedroom (or barn?). I think I have a good guess as to who it might be, but the genetic distance isn't close enough to convince most students of genealogy. But we are talking 300 years back to the event and most of the mutations are in the more rapidly changing markers. I don't follow what you mean by a how a NPE might strengthened the family line.

Anyhow, I appreciate all of the responses and I think this thread has pretty well accomplished what was intended for it. I particularly find Ann Turner's analogy with the laddar very useful.

I know all of us start out as novices and have to learn a lot. I don't want to bore forum members with my own problem, but the way it appears, there were a bunch of NPE's happening back in that late Colonial - early post-Revolutionary War period. My problem might be a good case study to present, and might help people from several family lines get past their brick walls. Does anyone think I should start a new thread with the details of my situation?

Thanks again for all of the help.

I hope I'm reading you right :I have a couple of thoughts 4U, why not upgrade your test to 67 markers that brings closer matches, also in early lines from England 1500-1600's cousins married, and that throws things off, they also sometimes married niece to uncle etc.

[josh w.]

Yes, a NPE. I guess I need to use the right terminology. It looks like a wife of someone in my family line let a man other than her husband into her bedroom (or barn?). I think I have a good guess as to who it might be, but the genetic distance isn't close enough to convince most students of genealogy. But we are talking 300 years back to the event and most of the mutations are in the more rapidly changing markers. I don't follow what you mean by a how a NPE might strengthened the family line.

Anyhow, I appreciate all of the responses and I think this thread has pretty well accomplished what was intended for it. I particularly find Ann Turner's analogy with the laddar very useful.

I know all of us start out as novices and have to learn a lot. I don't want to bore forum members with my own problem, but the way it appears, there were a bunch of NPE's happening back in that late Colonial - early post-Revolutionary War period. My problem might be a good case study to present, and might help people from several family lines get past their brick walls. Does anyone think I should start a new thread with the details of my situation?

Thanks again for all of the help.