scatter error

[cfz1998]

Hi! @zkamvar

According to the tutorial for Discriminant Analysis of Principal Components (DAPC). The scatter does not work well.

library(adegenet)
library(vcfR)

x <- read.vcfR("my.vcf", verbose=F)
y <- vcfR2genind(x)

grp <- find.clusters(y, max.n.clust=100, n.clust = 3)
# 40 for PCA
dapc1 <- dapc(y, grp$grp, n.pca = 40, n.da = 4)
# scatter(dapc1)

scatter(dapc1)

May be the big number of pca.cent?

Thank you for your reply!

[zkamvar]

grp <- find.clusters(y, max.n.clust=100, n.clust = 3)
# 40 for PCA
dapc1 <- dapc(y, grp$grp, n.pca = 40, n.da = 4)
# scatter(dapc1)

You are using the same number of Principle Components to train your model as you used to detect your clusters. You are overfitting the model. Reduce the number of principle components.

[cfz1998]

Hi! @zkamvar I got the same result when i reduced the number of principal components to detect clusters.

[zkamvar]

It's not an error. This means that you are comparing groups that are quite disparate.