Sunday, March 3, 2019

MyHeritage Theory of Family Relativity™ is Imperfect - Be Careful!

MyHeritage announced their "Theory of Family Relativity" feature last Wednesday - see   Introducing The Theory of Family Relativity™ — a Genealogy Game-Changer.  There is lots of demonstration and explanation in their blog post. 

I wrote Working With the MyHeritage Theory of Family Relativity™ yesterday and walked my readers through one of my successful "Theories."

1)  In the latter post, I mentioned that I had 15 DNA Matches (out of about 6,126 at this time), only about 0.25% of my DNA matches.  Don't expect that this feature will help you find the common ancestors of all of your DNA matches.  More "Theories" will be found when more genealogy work is done - and more MyHeritageDNA users develop a well-populated tree.

2)  A user needs to evaluate every relationship postulated in these "Theories."  Many "Theories" are comprised of one to many comparisons between trees and other trees or records to connect you with a DNA Match using their tools.  MyHeritage tries to help with a "Review" button that shows your data for the two connecting persons at each stage of the "Theory," and the Review button comes with a confidence estimate, in percentages, that evaluates the likelihood that the comparison is accurate.  

Out of my 15 "Theories," I think that nine of them are Correct, and six of them are Incorrect.  Of course, I'm comparing them to my own MyHeritage tree profiles, but I have a great deal of confidence that my tree profiles are correct.  

3)  Here is the "worst" of my 15 "Theory" connections with my DNA Matches.  This "Theory of Family Relativity" connected me to a 5th cousin DNA Match (whose name I've hidden for privacy reasons).  

a)  Here is the "Theory" chart showing the relationship with only the relationship results shown (not the connections):



As you can see, my line on the left goes up through the Seaver line to my grandfather, then to a Chase female, then to an unknown female, then to a Sarah Pound  (1785-1857), to the common ancestors, John Pound (1735-1790) married to a Sarah Martin (1737-1825).  On my DNA Match's side of the chart it goes from the same John Pound (1735-1790) married to Sarah Kester (1767-1845), with daughter Nancy Pound (1788-1862), who is apparently Sarah Pound's half-sister.

The "Unknown" person in the middle of my line concerns me.  What really concerns me is that the mother for my grandfather, Frederick Walton Seaver (1876-1942) is incorrect.  His mother was Hattie Louisa Hildreth (1857-1920), not Melissa Mary Chase (1849-1916).  How did that happen?

b)  I clicked on the "See Full Theory" button at the top right of the screen above, and saw my half of the "Theory of Family Relativity" for this DNA Match:


This "Theory" is comprised of nine different MyHeritage family trees, none of the mine.  My name (in the lower left), comes from someone else's MyHeritage tree.  This tree owner #1 only added me and my father to her tree.  So the "Theory" then matched my father with my father in a second MyHeritage tree.  MyHeritage assessed that match as 60% confidence (the green button circled in red on the screen above).  Why only 60%?  Probably because the trees had incomplete information for the person).  This tree owner #2 had only my father and grandfather in their tree.  The "Theory" then found a "Fred Sever" in a tree #3 with the same birth and death years, and assessed that as 20% confidence.  That tree #3 had Melissa Mary Chase as the mother of "Fred Sever" who was then matched with the "Unknown" person in tree #4 (a private tree, so the person shown is deceased but private), and that was assessed as 100% confidence.  The fifth connection on my side of the match is 100% confident according to the "Theory."  If I multiply all of the confidence numbers together just on my side of the tree, I get 12%.

MyHeritage offers a Confidence level of 20% for the whole Theory connection from me to my DNA Match.  Is that because it is the lowest number on the chart?  In my opinion, the overall Confidence level should be a multiplication of every confidence level on the chart - in this case it would be 9%.

The right answer, in my humble opinion, is 0% because of errors and limited information in the trees used to construct this Theory..

c)  Here is my MyHeritage family tree pedigree chart showing my Seaver line back many generations.


There's my grandfather, Frederick Walton Seaver (1876-1942) in my tree, son of Frank Walton Seaver and Harriet Louisa Hildreth.

d)  I went back and looked at the Review screen for my grandfather (in Tree #2) with "Fred Sever" in Tree #3, and saw:


Now I see why the comparison is only 20% confidence.  Tree #2 has only a name (which doesn't match exactly), a birth year (which does match the other birth year), and the death year (which does match the other death year).  Tree #2 has no parents and no siblings, while Tree #3 has parents and siblings.  Tree #2 has a wife and child.  Tree #3 has no wife or child.  20%!!  

4)  I understand how this can happen...I just showed you how it happened in this case.  I'm an experienced genealogist with a very well-populated tree, so my error-detector went off as soon as I saw this "Theory."  Other MyHeritage tree owners may not have an error-detector at all - they may trust what MyHeritage tells them.  

5)  How can a real bad error like this be corrected?

a) First, only my tree should be used to start off my side of the "Theory."  If someone doesn't have a tree, then too bad - they won't get any Theories to help them with their DNA Matches.  MyHeritage should not be taking my tree information from someone else's tree.  Of course, if they used my tree, they would not have found this particular theory.

b) Second, I think there should be some sort of overall confidence screen on each side of the "Theory" screen.  In this case, my side was 12%.  My opinion is that a confidence of less than 50% on either side should not be shown as a "Theory."  

c) Third, I think that nine "comparisons," especially of MyHeritage family trees with limited information for some comparisons, is a bit much for any "Theory."  

d)  Fourth, I think that the larger Comparson "Theory" chart with the Review buttons should be shown the user before the Relationship chart.  Many users will see the Relationship chart, think "well, MyHeritage shows it to me, it must be right!"

6)  But, and it's a BIG BUT - this DNA Match shares 8.6 cM DNA with me.  Not a big number.  Certainly it is in the 3rd to 5th cousin range.  I can't tell if this is a "real" DNA Match or not because I have not looked at the chromosome browser (and I'm not sure that would help!).  I have not looked at the MyHeritage tree of the DNA Match person to see if I can figure out where the relationship is, but I will.

7)  The "lesson learned" here is "You have to critically evaluate every relationship shown in the Theory of Family Relativity before you accept the connection to your DNA Matches as accurate.  You cannot ASSUME that the Theory is accurate."  

8)  I don't know what other users percentage of correct "Theories" is, but I know that mine is only 60% (9 out of 15).  I hope yours is better than mine.

9)  Every new feature on every website has problems after the initial release.  My hope is that MyHeritage will fine tune this "Theory of Family Relativity" to provide more accurate Theories. 


===================================================

Disclosure:  I have a complementary subscription to MyHeritage, and have received material considerations from MyHeritage over the past ten years.

The URL for this post is:  https://www.geneamusings.com/2019/03/myheritage-theory-of-family-relativity.html

Copyright (c) 2019, Randall J. Seaver

Please comment on this post on the website by clicking the URL above and then the "Comments" link at the bottom of each post.  Share it on Twitter, Facebook, or Pinterest using the icons below.  Or contact me by email at randy.seaver@gmail.com.

2 comments:

Doris Wheeler said...

I'm happy to report that, while I have only four on this Theory of Relativity list, they are all accurate, and only one of the four was known to me before.

cathyd said...

I just have 2 theories on my own DNA results; my mom has 10 theories. I noticed, as you said and also as Ancestry does in Thru Lines, MyHeritage is pulling from myriad trees. Ugh. Also, it's pulling from Geni.com and FamilySearch.org That is not so great either. For example, my mom's theories were mostly close to 100% accurate. But for those theories, the data was pulled from Geni or FamilySearch AND it was my own data!! While I think I am correct for this particular branch of my mother's tree, I am far from infallible. That the data used is not truly independent, to my mind makes these particular Theories of Relativity useless.