When are DNA Connections Stronger?

Question

I did a little experiment with my 7th great grandfather Joel Halbert. On his page there are 54 DNA Connections listed. I wondered if those who are descended from the same line would have more matches from that line.

So I grabbed all of those with a listed GedMatch number and came up with 24 in total that worked and weren't listed as private. Of the 7 listed children of Joel Halbert were the following number of connections counting myself; Sarah (2), John(3), Hannah(8), William(3), Martha(5), Joel(2), and Frances(1).

My assumption was that there would be more matches between those from the same line as those between different lines.

Here are the results. Out of 50 matches total there were 18 DNA matches (7cm or higher) between those in the same line and 32 DNA matches between those in different lines. So only 36% were from the same while 64% were from a different line.

Some conclusions and limitations. First I was surprised that there weren't more matches within the same lines. But the fact that were more matches across lines means that there is more hope for triangulation. Second, I used my phased DNA from my paternal line for this but first entered my unphased kit and noticed that I had more matches. This means that unfortunately, most if not all of the other kits were probably unphased and are likely matching DNA on the other side of the family and not just because they are descendents of Joel Halbert.

Ah, this stuff really is complicated isn't it?

Answer 1

Here's another way to look at it.

Since there are 7 family lines, if it were equally likely to have a match, regardless of the family line, than only 1 in 7 of those matches would be from the same line. That's 14%.

Your 36% is 2.6 times 14%, so your results are telling you that being in the same line make it 2.6 times more likely to have a match. That seems pretty reasonable.

I was recently looking at similar statistics within my brother's DNA matches on AncestryDNA, except for descendants from a pair of gt-gt-gt grandparents, not 7gt. He has 9+12+3+18+10+2=54 matches to people on the other branches (4Cs, give or take a generation) and 17 matches to people on our branch (3Cs, give or take a generation).

My situation is different, of course - I don't consider matches between people from the other branches (and I could only gather that info for matches over 20cM, if I tried), and there are undoubtedly DNA tests for descendants that my brother doesn't match, who are not included.

However, I think the basic underlying reason is the same. With 7 branches of the family, there are just a lot more people who are 4Cs to each other than there are people who are 3Cs. Even if the 3Cs are twice as likely to match, it's not enough to offset how there are 6 times as many 4Cs out there.

Comments

We have the simple doubling each generation to account for the number of ancestors we have (i.e. 2,4,8,16...) But even that is not perfect (as when cousins marry). But going down from an ancestor the math to figure out how many cousins per generation gets a whole lot more sketchy. There are baby booms and busts, childless couples, and then to make matters worse - only a percentage of those are represented on Wikipedia, and only a few of those have a gedmatch account. Is there any rough calculation of how many more 4C than 3C there are, or between any generation for that matter?

---

As for estimation of cousin numbers, Jonathan, that can be extremely difficult to generalize, as you noted, and will also see geographic populations and time period fluctuations. One possible resource that attempted that sort of count was a paper by Henn, et al., published in PLOS One. See in particular Table 2 under "Results," and the brief discussion under "Methods."