[adegenet-forum] Scaling in a combined Principal Component Analysis of morphological and genetic data

[Crameri Simon]

Hi Thibaut

I’ve got a question about how to best scale genetic and morphometric data for a combined Principal Component Analysis of both datasets using R/adegenet. Scaling of genetic data has already been a topic on the adegenet forum, but I think not in combination with morphological data.

Here is where I am unsure: I'm using dfssr <- adegenet::df2genind(datassr) to create an allele table of microsatellite data, then I use dfcombi <- cbind(dfssr at tab, dfmorpho) to combine the genetic data with morphometric data of the same diploid samples. Now I want to use PCA(dfcombi) to analyze the combined genetic and morphometric data in an integrative way. Scaling of the morphometric variables is important since the morphometric dataset contains variables measured on very different units, and variables with large variances would load the principal components too heavily compared to other, perhaps important variables with lower absolute variance. By contrast, scaling of the allele table (coded as 0, 1, 2 as in the @tab slot of genind objects) will (at least I think) lead to the principal components being more loaded by very rare alleles and very common alleles (i.e. those variables with low variance).

Since you have a lot of experience with multivariate genetic data, what would you suggest?

1) Scaling of the morphometric data only and cbind it with the unscaled allele table to perform PCA(dfcombi.dfmorpho.already.scaled, scale.unit = F)?
2) Scaling of the morphometric data and the allele table, i.e. PCA(dfcombi, scale.unit = T)?

The goal of the integrated PC analysis is to find entities that represent single species in a multispecies dataset.


Thank you in advance
Simon


**********************************************
Simon Crameri

ETH Zurich
Ph.D. candidate
Plant Ecological Genetics
Institute of Integrative Biology (IBZ)
Department of Environmental Systems Science
Universitätstrasse 16, CHN G27
8092 Zurich, Switzerland

E-mail: simon.crameri at env.ethz.ch<mailto:simon.crameri at env.ethz.ch>
<mailto:niklaus.zemp at env.ethz.ch>Web: www.peg.ethz.ch<http://www.peg.ethz.ch/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/adegenet-forum/attachments/20160413/11ac0e08/attachment.html>