search

JulienGAMartin / pedtricks

reboot of pedantics R package
https://juliengamartin.github.io/pedtricks/
Other
1 stars 0 forks source link

pairwise relatedness metrics #18

Open joelpick opened 1 week ago

joelpick commented 1 week ago

I was getting different values for the pairwise relatedness metrics with the code I had written to the pedantics code. I've worked out the problem, and I think the pedantics metrics are a bit off. So for the cumulative relatedness, lots of small bins are made. The bins are then labelled with the midpoint of the bin, not the maximum. So the name of the bin is NOT the cumulative proportion to that relatedness, but to a slightly higher one. To get the pairwise metrics in the output, these cumulative metrics are then reported, which are slightly higher than I think they should be.

If people agree I will correct that.

JulienGAMartin commented 1 week ago

When changing the code to use the sparse matrix for A, I remember thinking that it was odd to use the mid-point of the bins instead of the maximum. I would agree that it can report slightly higher metric than expected. I did not had time to do it and should have created and issue.

So I agree and thanks for fixing it.