We should compare with VCF, BCF and Savvy (I think SAIGE supports this), using the simulated data we already have. We don't have to go up to 1M samples, just whatever seems reasonable. We can plot the time to run the association test based on traits simulated by tstrait against increasing sample size, as usual.
Based on this code: https://github.com/Will-Tyler/SAIGE/pull/1
We should compare with VCF, BCF and Savvy (I think SAIGE supports this), using the simulated data we already have. We don't have to go up to 1M samples, just whatever seems reasonable. We can plot the time to run the association test based on traits simulated by tstrait against increasing sample size, as usual.