How should I compare my pi results from pixy with studies that use other approaches? about pixy HOT 2 CLOSED

Hi Tal,

Thanks for your question, it is a perfectly reasonable one! The similarity between pixy's estimates and estimators that assume missing sites are homozygous for the reference allele is a function of the amount of missing data. The more missing data there are, the more biased the estimates of pi and dxy become. In cases where missing data is very high (e.g. computing over windows using RADseq or similar type data), pixy will differ a great deal from other estimators.

If one wanted to standardize estimates post hoc, the key quantity would be the number of sites in the window of interest that had sufficient depth to call a genotype in the focal individuals (i.e. those that would have appeared as invariant sites in an all-sites VCFs). Silas Tittes made a really nice command-line program for computing this quantity from bam files https://github.com/RILAB/mop. You would need access to the original bam files to compute it, but that is unfortunately the only way I am aware of doing this!

I hope that helps. This is unfortunately a very common problem and there isn't really a great solution for it other than encouraging better practices!

from pixy.

Thanks Kieran. That's what I was expecting I suppose :). Glad to see at least that more people are using pixy in publications over the last year. I will check out the Mop program.

from pixy.