Age estimates for FGC23343 subclades

[benowicz]

I recently received results for a BigY700 kit I manage for a relative. Comparing these results to block trees at FTDNA, the Big Tree and YFull, I'm staring to get an appreciation for some of the complexities in estimating Most Recent Common Ancestors ("MRCAs"). I think the biggest problem for outside projects is ensuring full comparability of kit resolutions, but even the testing companies themselves seem to sometimes get stymied by no-calls, etc. One universal problem that I don't think is going to be simply and definitively resolved any time soon is arriving at an average mutation rate. Unless we're talking about a relatively recent MRCA with a large number of matching donors, age estimates are necessarily going to remain somewhat speculative.

FGC23343 has yet another weird problem that I'm not sure I completely understand. One subclade (i.e., FGC28370) has an abnormally large number of mutations recorded within its defining blocks--like twice as many as the next largest branch. How could this happen? I can only guess that it has something to do with a difference in test resolution among individual kits, but the magnitude still seems astounding to me. I know there was one SNP (i.e., FGC62822) that was recently allocated out of FGC28370 to FGC23343, probably due to some kind of low-coverage induced no-call situation, but I can't imagine too many more such "corrections" to the tree in the near future.

But I think I've come up with a solution to these dating problem that produces some pretty reasonable estimates. Below are my current guesses, based on a normalization algorithm that adjusts the reported number of SNPs based on the average number of SNPs among the lineages under FGC23343, and an assumed 85 year mutation rate.

Est. birth year	Clade
250 A.D.	FGC23343
304 A.D.	FT372222
537 A.D.	BY97678
1430 A.D.	FGC28370
1462 A.D.	Y74676
1647 A.D.	FGC28369

Luckily, some of these dates are recent enough, and supported by STR profiles so as to allow some sort of tests to corroborate. I haven't yet exhausted all such possibilities, but preliminarily I can say that the mid-points of my normalized SNP-based curve and the STR curve for FGC28370's MRCA are only 30 years apart, per the McGee Y Utility and the McDonald Y DNA calculator. The confidence intervals for some of these estimates can be pretty shocking, but the MRCA dates seem to hold together with a coherence that I haven't seen in any other estimates. Seems like a good step forward in analyzing a tricky case.

[benowicz]

This selection from a Wikipedia overview of the history of the Vascones tribe puts some of these dates in an interesting context. At just about the same point in time that Adolf Schulten suggested the Vascones were expanding from Navarre west into Spain and north into Aquitaine, FGC23343 and its earliest clades are spreading into precisely these areas.


"During this period, after the time of Ptolemy and contemporary to the times of instability caused by the Germanic invasions, the documents about the Vascones and other tribes of the northern Iberian Peninsula are scarce, and as a result there is little information about the Vascones during this time.

The Visigothic Kingdom circa 560. The Vascones and Varduli in the north.

The chronicler John of Biclaro (c. 540 - after 621) mentions the Vascones in a story about the foundation of the city of Victoriacum by the Visigoth king Liuvigild^[13] and Gregory of Tours (538-594) mentions the incursions of Wascones in Aquitaine during the year 587.^[14] From these extracts and being the neighboring tribes absent in the historiography, Adolf Schulten (1870-1960) proposed the theory according to which at some point between the mid-2nd Century and late 4th Century, took place an enlargement of the territory of the Vascones, first in the west, occupying the lands of the Caristii, Varduli and Autrigones,^[15] and later in the north in Aquitaine.^{[16][full citation needed]} Schulten considers this to be the reason for the adoption of the name Gascony, which derivates from Gascon, which comes from Vascon, and used to denominate a region that includes the present-day Northern Basque Country.

Claudio Sánchez Albornoz, Spanish historian, (1893-1984) on his work "Los vascones vasconizan la depresión vasca" (The Vascones "basquize" the Basque depression) published in 1972 enlarged this hypothesis, relying on linguistic analysis: when invading the territories of what today is Biscay, Gipuzkoa and Álava displaced to Castile part of the Caristii, Varduli and Autrigones, who took refuge in the mountains; the ones who had not been displaced were "Basquized",^[17] while perhaps the Caristii, Varduli and Autrigones already spoke languages similar or related to the Basque language.

However, research during last decades has called into question the possibility of an expansion northwards (J.J. Larrea). The inroad of the Vascones onto the plains of Aquitaine in 587 seems to be short-lived—they make their way back to the mountains—and archaeological findings in Eauze or Auch do not reveal instability or destruction during the alleged expanding period up to the mid-7th century. Another theory suggests a contemporary identification made by the Goths and the Franks of the Vascones (the most dynamic tribe) with all Basque speaking, Basque-related, or non-Romanized tribes."

[benowicz]

After being given this 2015 technical paper, I was able to construct a more scientifically robust aging analysis of these subclades.

https://forums.familytreedna.com/for...202#post329202

I don't have direct access to data for all the relevant variables, but I can make some pretty good educated guesses, and as a result, the mean predicted values and standard deviations shifted a bit for some of the clades.

Est. birth year	Clade
463 A.D.	FGC23343
559 A.D.	FT372222
861 A.D.	BY97678

The thing is that the predictions are specific to the array of observed private variants and the established predictions for parent clades, even the level of coverage achieved by specific testing products. So while I can't really offer any off-the-rack estimates of mutation rates, I will say that in these examples, average mutation rates ranged from 62 years per SNP to 74, significantly lower than the subjectively determined numbers that had been recommended to me informally.

The confidence intervals extend well into the so-called Viking era, and that really interests me given the information surfacing as to the likely origins of FGC28370 and FT372222 in western Normandy. There is plenty of contemporary documentary evidence that the viking bands active on the Loire and Garonne rivers picked up some local collaborators along the way, and perhaps FGC23343 and its subclades are further evidence, given the almost certain origin of the parent clade Z209 in the Basque country.

[benowicz]

. . . The thing is that the predictions are specific to the array of observed private variants and the established predictions for parent clades, even the level of coverage achieved by specific testing products. So while I can't really offer any off-the-rack estimates of mutation rates, I will say that in these examples, average mutation rates ranged from 62 years per SNP to 74, significantly lower than the subjectively determined numbers that had been recommended to me informally. . . .

Just to be clear, the process is a very objective, rule-governed procedure. At no point does a completely random estimate enter the process. It's just that I don't always know what platform (i.e., Big Y 700, Big Y, etc., etc.) a given donor tested under. The coverage typically achieved by the individual products varies significantly. Nor do I always have enough information to work out definitively how many private variants a specific donor has--FTDNA aggregates the reporting of these under their most recent block, although I can usually work it out algebraically by reference to data from public projects.

So my inability to recommend a single useful rule-of-thumb about typical mutation rates is about the adherence to scientific principles rather than a relaxation of discipline. Quite the opposite. In that light, these most recent figures should be considered conservative--it is likely that the TMRCA is somewhat more recent, but not equally likely that it is somewhat more remote.

I'm just stunned at how useless these rules-of-thumb turn out to be. There's an incredible amount of variation in mutation rates among individual donors--but there is only one TMRCA estimate that resolves them into a single, most plausible relationship. I can objectively associate probabilities to lots of different TMRCA estimates, but only by reference to this single, most plausible estimate.