How to calculate Nucleotide diversity (Pi) and Heterozygosity confidence intervals?

[jezk...@unlv.nevada.edu]

Hi,

We are trying to compare Pi and Heterozygosity between two populations. We predicted that one population should exhibit lower values of both indices because it was established only recently via a stepping stone range expansion. We have ca. 10 individuals in each population and 5-10k SNPs (depending on the threshold used in "Populations" in Stacks). Our mean values are consistent with our predictions. For example, for one of our thresholds (10x, 0.75), Pi for one population is ca. 0.0046, for the other one 0.0039. My question is, how do we assess that the two populations exhibit significantly different Pi? We first used variance provided by Stacks but the variance values are very large. If we understand the calculation of the variance correctly, it is because the variance represents variability across individual loci and not a confidence interval around the mean value. So we tried to bootstrapped the values from the <batch>.sumstats.tsv file. The resulting values are super small, around 0.00005 or so. My questions are following:

1) Is our logic correct in not using the variance values from Stacks as our confidence intervals to assess the significant difference in Pi and Het in our two populations?

2) Does the bootstrapping of the mean values make sense?

3) Do you have any other suggestions about how to assess whether two populations have significantly different Pi and Heterozygosity?

I will be very grateful for any suggestion or advice.

Thanks so much, Tereza