[adegenet-forum] Mantel tests and scaleGen

[Sam Fellows]

Hi,

Thanks for having this up as a resource. I'm not seeing the full answer to
my question in the archives, so I figured I'd go ahead.

I am testing a dataset of "unlinked" SNPs (generated by RADseq) for
isolation-by-distance effects and see that in tutorials available which
include mantel.randtest it's recommended that the genetic data are scaled
with scaleGen() prior to the Mantel test. However, I'm not seeing a
published example of why this would be recommended for this analysis. A
previous question asked had an answer that seemed to suggest rare alleles
might affect sPCA, but I'm unclear why this was suggested for a Mantel
test. I also noted a previous question where someone recommended not to
scale genetic data in general, but whether or not that applies to my data
is unclear to me.

Scaling my genetic data prior to running the Mantel test results in no
significant relationship (except in a dataset with a relatively large
amount of missing data) while leaving the data unscaled results in
significant (p < 0.05) relationships regardless of filtering.

In both cases, kernel density estimation plots of the pairwise distances
shows some somewhat messy clustering, and spatial PCA via multispati
(adespatial, wherein I leave the data unscaled) reveals some clear
clustering and some of what is likely IBD.

Any insight as to why a genetic dataset might be scaled/left unscaled for a
Mantel test would be helpful.

Best, and thank you in advance,
Sam Fellows

-- 
Sam Fellows
PhD Student/Teaching Assistant
Quantitative Systems Biology- Evolutionary Biology
University of California- Merced
Edwards Lab <http://danedwardslab.com/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/adegenet-forum/attachments/20200515/45425dd6/attachment.html>