There is a good method for calculating TMRCA that is not tripped up by population history, nor does it make any "continuum" assumption about the mutation process. Walsh describes the starting point for the method in the Stepwise Mutation Model section in his paper published in "Genetics 158: 897–912 (June 2001)". Contrast that method with other methods for computing TMRCA. The Average-Squared-Distance (ASD), "Variance" method, and variations on that method for computing the TMRCA, explicitly do a sum over the sample population - and simulations show that that reliance on the surviving population will lead to a gross underestimate of the true TMRCA. Part of the reason is to do with sons being born with a Poisson distribution (that is, there is a certain probability of a man having 0 sons, another probability for him to have 1 son, and a different probability for him to have 2 sons etc, with the probabilities drawn from a Poisson distribution). That process means that the vast majority of male lines will necessarily go extinct. But correspondingly, a small number of initial male lines will still survive and dominate. That is a feature of the Galton–Watson process, which is best known in the context of surname extinction, where in the long run all hereditary surnames will in theory go extinct except for a small number of surviving surnames. (It is happening with English hereditary surnames, and particularly so with Vietnamese hereditary surnames.) That is the first problem with any ASD or Variance based method for computing the TMRCA. It is tripped up by population history, and the chance survival of some lines over other lines that go extinct. A second problem with any ASD or Variance based method, is that those and related methods implicitly assume that an STR allele value (with example allele values being say DYS393=13 or DYS393=14) is a "continuous" variable, where in reality STR alleles can only ever have discrete integer values. For fast mutating markers, and over a long time frame, that assumption wouldn't make too much difference, but for a slow mutating marker or for a short time frame it makes a big difference. Some published papers only applied such methods to problems where the TMRCA was very big (and the "continuous" allele assumption was not such an issue), alternatively they restricted themselves to only fast mutating markers. Those published papers still had the problem of population history to handle though. I do note that Ken may have slightly mitigated the negative effect of the "continuum" assumption by replacing all STR mutation rates with just the average rate in his method. That would have the effect of down-weighting the contribution of the slow mutating markers. That is a loss of useful information, but the average rate may have the side effect of reducing somewhat the bigger problem of the "continuum" assumption when applied to a relatively young haplogroup such as I1. Nevertheless, Ken's TMRCA estimate for I1 is still way off compared to the method that both correctly factors in the discrete integer nature of STR allele values, and is mostly insensitive to population history. For me, the focus is really on the quality of the STR mutation rates we need to use to compute the TMRCA. I have been happy enough to use the Chandler rates for the first 37 markers, but the rates out to 67 markers I am not so certain about. Then there is the recent paper by Burgarella, which gives some new STR mutation rates (with confidence intervals) for 110 markers. Both I and Ken, have used the same mutation rates as input into our respective methods - although as mentioned, Ken uses a single average rate applied to all STR markers. In the Walsh formulation, which also computes the probability distribution of the TMRCA, each STR marker gets to use it's own mutation rate, whether it be fast or slow. Getting quality STR mutation rates out to 110 markers will be important. Finally, I should say that there will always be a high margin of error when computing the TMRCA using a limited number of STR markers with a non-perfect understanding of the mutation process for any given marker. Assumptions are everything. But using the nice (and wrong) uniform population history assumption, and the continuum assumption for STR allele values, are two assumptions that we don't need to make. Terry Footnote: For the physicists. If you were to think of the evolving STR allele values as the random drifting of a particle (Brownian motion or whatever), then under the "continuum" assumption, such a particle will move via diffusion to a position that has an expectation distance from the starting point that is proportional to the square-root of the time that has elapsed. Or squaring things, you could say that the average squared distance from the starting point is proportional to time. And the distribution of the final positions is a Gaussian. But that is under the "continuum" assumption, where the particle can make arbitrarily small steps. Under the "discrete" assumption, where a particle is on lattice and only has the option of staying at a lattice point or jumping a discrete integer amount to a new lattice point, then the (stochastic) formula for that motion is different. It is well-known for Brownian motion on a lattice, that the distribution of the final positions involves a modified Bessel function. For the Y-chromosome, the changing STR allele values over time, are similar to the "discrete" assumption above. The "continuum" assumption is only an approximation to that, which may not always be applicable.