[adegenet-forum] Request of DAPC analysis

[Das, Roma (ICRISAT-IN)]

Hello everyone,

*         Based on DAPC analysis, I am not sure whether I should treat the final group for individuals line as 1) prior group from find.clusters () or 2) group with maximum posterior probability after xval.DAPC(). As in scatterplot what we get after analysis, individuals are plotted based on prior group. Please advise

*         Is there a way to choose optimum number of discriminating functions to be used

Regards,
Roma

From: Das, Roma (ICRISAT-IN)
Sent: 28 August 2019 04:21
To: 'adegenet-forum at lists.r-forge.r-project.org' <adegenet-forum at lists.r-forge.r-project.org>
Subject: Request of DAPC analysis

Hello Everyone,
I am using DAPC using adegenet package for cluster analysis. However I am not sure if I am following the correct way to select n.pca and n.clust based on cross-validation.

I am following below steps


1.       I am using a genind object

2.       Used find.clusters() grp <- find.clusters() and interactively chose n.pca and n.clust. Based on plot, I selected n.pca=200 and n.clust=21

3.       Next used xvalDapc() to get some idea about number of PCs
xval <- xvalDapc(tab(fdat, NA.method = "mean"), grp$grp, n.pca.max = 300, n.rep = 30)

4.       Based on number of PCs achieving highest mean success and lowest MSE, I selected n.pca=50

5.       Further, I tried to narrowed the search of PC's with n.pca = 30:60
xval_optimum <- xvalDapc(tab(fdat, NA.method = "mean"), grp$grp, n.pca = 30:60, n.rep = 100,parallel = "multicore", ncpus = 6L )

6.       Finally I selected n.pca=30 based on number of PCs achieving highest mean success and lowest MSE from xval_optimum

My questions are:

7.       From cross-validation, it seems the optimum number of PCs is 30. Should I re-run find.clusters() with n.pca=30 and select n.clust interactively from plot

8.       And then re-run dapc() with n.pca=30 and output of n.clust from step 6. Please advise

Thanks,
Roma
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/adegenet-forum/attachments/20191020/1b35d93d/attachment.html>