I posted notes on the Wahlund effect and F-statistics a while ago. I’ve now posted an R Shiny application to illustrate the difference between Nei’s GST and Weir and Cockerham’s FST. The application simulates a sample of 25 diploid genotypes from 10 different populations. The genotypes are a multinomial sample from genotype frequencies calculated from Hardy-Weinberg expectations within each population, given the population allele frequency. That’s statistical sampling. The allele frequencies in each population are sampled from a Beta distribution with a mean of p = 0.5 and a variance of FSTp(1-p). That’s evolutionary sampling (or genetic sampling). Just as the individuals we sampled within each population are a sample of all individuals we could have sampled, the populations we sampled are a sample of all populations we could have sampled.
If you keep the parametric FST the same and just keep hitting “Go”, you’ll see that the genotype counts change every time. That’s the evolutionary sampling. You’ll find a link to the application on the lecture detail page, or you can link directly to the application on shinyapps.io.
As a reminder, if you’re interested in the source code for this or other R Shiny applications I develop for this course, they’ll all be available on Github.