Effect of non-random sampling on the estimation of parameters in population genetics

The amount and pattern of genetic variation in a population can be estimated from genes or DNA sequences sampled from the population. Although random sampling is assumed in almost all cases, we often do not know whether sampling is random or not. Using a simple non-random sampling model, the effects of non-random sampling on the estimation of parameters in population genetics were investigated. This non-random sampling model assumes that n genes are randomly sampled with replacement from m genes which were randomly sampled from a large random mating population, and various degrees of non-randomness can be generated by changing the value of m. The results obtained show that the effect of non-random sampling on the number of alleles and the number of segregating sites is substantially large whereas the effect of non-random sampling on heterozygosity and the average number of nucleotide differences is negligibly small unless non-randomness is extremely large. The effects of non-random sampling on the tests of neutrality were also investigated, and the results obtained indicate that the effect of non-random sampling is stronger on Fu and Li's tests than on Tajima's test.