[adegenet-forum] Reagarding sPCA: retained scores

[Jombart, Thibaut]

Dear Roy, 

scannf=FALSE allows the user to specify the number of retained factors (nfposi, nfnega), so this is no automated procedure.

Multivariate analysis is about identifying structures in the data, that is, separating interesting patterns from non-interpretable/irrelevant variability. If you keep 100 principal components, probably only the first few will be relevant and interpretable. Selection of PCs has been extensively discussed in the literature, and there is no automated method which works in every case (or actually, which works at all). In the case of sPCA the screeplot gives you hints, and the associated tests (global.rtest, local.rtest) can also be helpful. Please also have a look at the sPCA vignette (vignette("adegenet-spca")).

Cheers

Thibaut



________________________________________
From: adegenet-forum-bounces at r-forge.wu-wien.ac.at [adegenet-forum-bounces at r-forge.wu-wien.ac.at] on behalf of Roy Francis Mathew [RoyFrancis.Mathew at agrsci.dk]
Sent: 10 November 2011 08:53
To: adegenet-forum at r-forge.wu-wien.ac.at
Subject: [adegenet-forum] Reagarding sPCA: retained scores

Hi,
I have a question regarding sPCA scores.
While performing an sPCA, the scores to be retained can be retained automatically (scannf=F) or interactively.
How does it decide how many to retain? And if I retain 1 or 3 or 6. I can still plot out just 1 or 2.
Is there any reason why only a few are retained? Why not just retain everything by default? Does this affect any downstream analysis?
Thanks,
Roy
_______________________________________________
adegenet-forum mailing list
adegenet-forum at lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/adegenet-forum