Monday, 3 October 2016

Will you find a cousin using a DNA test?

Yes; likely many.

That's unless all your parents, grandparents, great-grandparents, great-great-grandparents and further back were their parent's only child in which case there are no corresponding cousins of that degree.

An article Cryptic Distant Relatives Are Common in Both Isolated and Cosmopolitan Genetic Samples includes a table reproduced at the ISOGG website under the heading How Many Cousins Do We have? The simplified version here shows that, despite the dramatic reduction in chances of detecting more distant cousins, results typically show more matches to more distant cousins. Simply, there are more of them.

The many more distant cousins connect on the many more lines down from shared ancestral couples, and in times when families were typically larger than today.

Just because cousins are detectable does not mean they'll be a DNA detected. To do so they have to also have taken a DNA test with results in the same database as your's.

The second table shows the probability of detecting at least one cousin for two different database sizes, 250 thousand and 2.5 million people tested. The calculation is based on the assumption the cousins will be found in a total population of 500 million encompassing the population of those who have likely tested, a subset of the 7.4 billion world population. If your near term ancestry links to the larger world population these probabilities will be less.
Size matters. The chances of detecting a fourth cousin in the larger database are more than four times those in the smaller.

Most genetic genealogists have the experience of having cousins as DNA matches but not being able to find the connecting recent common ancestors. We usually look for surnames in common in both family trees. But our genealogical brick walls mean we know a smaller fraction of surnames in the more distant generations.
The third table extends the second by including two additional columns. Column five shows the estimated probability of recognising the common ancestral surname of a cousin of specified degree. That depends on both you and the matching person identifying the surname in their respective ancestral trees. The final column combines the probabilities in the previous two columns to estimate the probability of both detecting and recognising the common ancestral surname of a cousin of specified degree, based on 2,500,000 person database.
There's a sweet spot around third and fourth cousin where you're likely to have the greatest chance of making a connection.
Additional factors not taken into account are identification of geographic locations in common and work on detailed segment analysis such as used by those searching for adoptee ancestry.
If looking for cousins the message from these estimates is to ensure your DNA result is in as large a database and as many databases as possible, and to include a family tree or ancestral surnames as far back as possible.


Flameseh? said...

Great from the statistical perspective! I've credited you over on the ISOGG FB page.

Linda Reid said...

I think these tables are very US-centric. More testees in the database are of more limited value to those who don't have American ancestry if the database is predominantly American. Many Americans do not know about their family lines before they became American so the odds become slim that there will be surnames of interest further back. Expectations have to be adjusted by place of ancestral abode.

Peter Calver said...

John, congratulations on an excellent post. Picking up Linda Reid's very relevant point, in an article on this topic a couple of months ago I used I used the estimates published by Ancestry DNA for the number of cousins of different degree (these take into account the rapid rate of organic growth in the British population during the 19th century). Things aren't quite so bad as it might at first seem for those of us with British ancestry, although the general point about surnames not being recognised remains, which is why in the same newsletter I recommended looking for geographical overlaps between trees.

Of course our 'brick walls' make the process more challenging, but for many of us the primary motivation in testing our atDNA is to knock down those very 'brick walls' so I for one wouldn't want it any other way!

Donna Fraser said...

John, I recently found that one of my friends in our small Qualicum Beach Family History Society is also a 2nd to 4th cousin per 23 and Me DNA. We are both researching our maternal lines in Scotland and had no idea we were related. Now the hunt is on to find our MRCA. I'm delighted to be her cousin! The thrill of the chase continues...