Thank you for your kind words! 😊 Happy you liked it!

Also, do not forget to check out our #rstats package implementing our whole framework, with a nice tutorial included:
cran.r-project.org/package=Reac...

As I mention in the end, we decided to support the @peercommunityin.bsky.social initiative, but it does mean that some people not monitoring this part of scientific publishing might miss our paper.

So, if you find it interesting, please share the word! 🙏

This method is strongly inspired by previous by O Ovaskainen and J Merila :
academic.oup.com/genetics/art...

The key differences are that we use a different test statistics and a quite generally applicable estimator of co-ancestry developed by J Goudet and B Weir:
journals.plos.org/plosgenetics...

We show in the paper that doing so results in calibrated tests of neutrality, even in the face of highly structured scenario, contrary to classical Qst-Fst comparison. 🎉

The clever trick is since both estimates (using between- or within-population co-ancestry) refer to the same ancestral VA, they should be equal under neutrality.

LogAV thus compare these estimates as the Log-ratio of Ancestral Variances (hence the name) to 0 (the expectation under neutrality).

Instead of summarising co-ancestry with a global Qst, or Fst, index, LogAV uses the full, carefully designed, matrices of between- and within-population co-ancestry to refer to the same ancestral additive genetic variance.

As such, it fully accounts for the actual population structure.

I guess it also depends on the type of article. You can have an LLM write up a literature review for you somewhat easily (this says nothing about the quality of the output...) for example.