Y-DNA TiP Report: Fundamentally Flawed Algorithm

[Rheiner]

Hi. I've been using the TiP report, as an easy yardstick for probabilities... but there's a glaring issue: haplogroup incompatibility.

The top name on my 67-marker match list computes much the same as the others on the list, despite the differences in mutation rate supposedly tracked by the algorithm.

Maybe so -- but it doesn't factor in the haplogroup at all! Our ancestors diverge thousands of years ago -- but since the match list only displays terminal SNP, there's no clue that there is no chance whatsoever that we are genealogically related. It's a waste of time to display "matches" from other haplogroups. And a bigger waste of time to have to google each terminal snp to see if it's worth considering.

This would be very easy to fix in your algorithm. Hence this gripe. Thanks.

[prairielad]

When referring to haplogroups, yDNA STR testing does not test yDNA SNPs to determine haplogroup.
FTDNA predicts a basic haplogroup based on ones STR marker signature (predicts an older parent branch of ones final subclade placement of haplogroup)

That said, people do additional testing to determine a deeper subclade branch of haplogroup.
For ones that have done yDNA SNP testing to determine a deeper subclade branch, the SNP tested positive for to determine the deeper subclade branch will be listed in the Terminal SNP column
Terminal SNP does not mean the SNP that determines one furthest known subclade branch of haplogroup (ones full haplogroup to the deepest known subclade branch of yDNA tree), it just displays how deep that match has tested to.

Most of your matches haplogroups, if you look them up in yDNA tree, will show that their haplogroup or yours will be a older parent branch of the other.
If all matches did full yDNA SNP testing (ie BigY) they would likely all end up in the same final subclade branch of haplogroup.

[Rheiner]

Thanks, but this doesn't seem on point.

To be specific: I am classified as R-Z367 at the moment, downstream from U-152. But I am matched with a host of others who are downstream from U-106, or L-21, and so on. They are in different subclades entirely.

On my match list, matches which display SNP (usu. Big Y results) are displayed -- even though they are downstream from U-106, or L-21, etc. Still they are ID'd only by SNP. So even though the match is bogus, I have to look it up to be sure.

Yes, there are lots of matches displayed as R-M269, etc. But if testers have had further testing, those should be factored into the algorithm. Now, they are not.

[bartarl260]

The TiP Report could probably use a tech-refresh, but you need to remember it was written as a tool to evaluate STR results, not SNP results.

It would be nice if they're do a BigY variant of the TiP Report(using the BigY SNPs and STRs), but I don't think they're at the point where they have a very high confidence level on mutation rates for things beyond Y111. Until that happens, I won't be holding my breath. This also ignores the whole matter of how you even go about determining SNP mutation rates, Yfull is pulling numbers out of an unmentionable place. Yes, there is some basis for their conclusions and estimates, but the science is still emerging on that front, while FTDNA does try to be on the leading edge of DNA testing, they also do try to stay away from the bleeding edge of analysis and conclusions, which is why a lot of people get annoyed by things like FTDNA not venturing very deep into haplotree predictions based on STR test results alone.

As it stands, the TiP report is comparing Apples(STR result panels) to Apples(STR Result panels), which is what it was implemented to do. You're basically complaining that it isn't comparing Apples to Oranges, and sorting them by specific sub-types along the way, and a slew of other options. Yes, those capabilities would be nice, but when you're talking BigY, the bulk of the User Base at present are better served by tools like the Block Viewer and the SNP match/mis-match list than they would be with an upgraded TiP report.

Yes, there are some BigY Users that could take advantage of an upgraded "Deluxe" version of the TiP Report which can identify the difference from an apple and an orange, but they're still a very small minority within FTDNA's user base. FTDNA is a business, they're going to focus on things which provide benefits to the most users, and currently the TiP report getting an upgrade isn't likely to be on that list just yet.

[John McCoy]

The "fundamental flaw" is that the TiP report does not consider all the available information. I do consider that lapse a fundamental one!

However, the notion of a random process (such as Y chromosome mutations) with a small sample size (a statistically small number of STR's and SNP's, as well as a small number of individuals that have been tested) that nevertheless keeps accurate time (the "molecular clock" idea) is itself very problematic, to the extent that we should probably not take the results of any TMRCA estimate as fact, but rather, as an educated guess. The probability estimates of the TiP report attempt to express this uncertainty, but I don't think we have sufficient data even to guess at the level of uncertainty.

[bartarl260]

The "fundamental flaw" is that the TiP report does not consider all the available information. I do consider that lapse a fundamental one!

However, the notion of a random process (such as Y chromosome mutations) with a small sample size (a statistically small number of STR's and SNP's, as well as a small number of individuals that have been tested) that nevertheless keeps accurate time (the "molecular clock" idea) is itself very problematic, to the extent that we should probably not take the results of any TMRCA estimate as fact, but rather, as an educated guess. The probability estimates of the TiP report attempt to express this uncertainty, but I don't think we have sufficient data even to guess at the level of uncertainty.

The big thing I look for when using the TiP calculator, particularly when able to ping a 67+ marker test result(which you can do on even a 12 marker match), is looking for 0% probability matches or very low probability ones. If it says there is next to no chance that you're related anytime in the last X number of generations, that probably isn't a line worth investigating at all.

Likewise, a high probability match in the Y67/Y111 range becomes one that might be worth active pursuit.

The stuff in the mid-range? Could be closely related, or very distantly so, probably not worth much active pursuit though until or unless you a reason that isn't the TiP report to try seeking the person out. Being GD2 with my father by Y37.TiP has a hard time believing I might share an male-line ancestor with my father anytime in the previous 4 generations, so your mileage can vary by rather a lot.