[adegenet-forum] yes/no answer. are xval results only relevant for the dapc part of dapc analysis and not find.cluster

[Thibaut Jombart]

Hi Ella,

1) Yes, only for DAPC, to prevent over-discrimination of the groups when
there are many variables

2) n.pca.max: there's no real rule, this is to avoid useless computational
times; the idea is to avoid i) retaining too few PCA axes, in which case
your DAPC will be sub-optimal in terms of discrimination, and ii) retaining
too many PCA axes, in which case discrimination will be artefactually good
on your data (but bad for cross validation, hence the use of it)

Hope this helps
Best
Thibaut

--
Dr Thibaut Jombart
Associate Professor in Outbreak Analytics, London School of Hygiene and
Tropical Medicine
Senior Lecturer in Genetic Analysis, Imperial College London
President of RECON: repidemicsconsortium.org
https://thibautjombart.netlify.com
Twitter: @TeebzR


On Fri, 25 Jan 2019 at 16:09, Ella Bowles <bowlese at gmail.com> wrote:

> Hello,
>
> I am writing to follow up on a couple of questions that were not answered
> on the forum. It should just be a quick yes/no answer.
>
> 1). I'm wondering if xval only needs to be used for the dapc part and not
> find.cluster?
> 2) Can I verify that for xval anaysis, n.pca.max should be set to n/3?
>
> I have 186 individuals spread over 11 sites for which I am trying to sort
> out the population structure.
>
> In case this thread can be correlated with my last on the forum, the
> previous question had the title "
> are xval results only relevant for the dapc part of dapc analysis and not
> find.cluster"
> With thanks,
> Ella
>
> --
> Ella Bowles, PhD
> Mitacs Elevate Postdoctoral Fellow,
> Department of Biology, Concordia University
> Niskamoon Corporation
>
> Website: https://ellabowlesphd.wordpress.com/
> Email: bowlese at gmail.com
> _______________________________________________
> 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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/adegenet-forum/attachments/20190125/832fa258/attachment.html>