Average Nv = 6.48

all stem from using models to describe actual populations as opposed to making idealized predictions. The model assumptions, which seem reasonable in the abstract, have become critical now when faced with a realistic sized data set. It reminds us again of the difference between a parameter and a parameter estimate. The data we used to make the estimates probably do not fit the model assumptions well.

Using only 10 replicate populations to estimate the effective population size is clearly not enough given that the Markov model assumed an ensemble population that approaches infinite size. In some cases the heterozygosity actually increased between generations, although the expectation in equation 3.52 is that it should only decrease. The average Ap was not zero even though with many replicate populations it is expected to be. The estimates also only utilized a single generation of genetic drift while the expectation for a fluctuating populations was over four generations. It seems that the change in allele frequencies and heterozygosity in the simulated populations were disconcertingly random! All this serves to remind us that assumptions like "many populations" or "over long time periods" may not be biologically realistic because a data set or organisms themselves may not play by the same rules. Violation of model assumptions often leads to poor or imprecise estimates of parameters, just as we have seen in this example.

0 0

Post a comment