Interpreting the BIC for optimum k-means cluster

[Jason Berg]

Hello,

I have been running DAPC analysis on a data set that includes 16 diploid and one triploid population of the genus Mimulus. The scatterplot shows a signal for 3-5 clearly defined groups (the diploid populations are clustered into three or four groups, and the triploid population is clearly separated from these diploid pops in its own group). However, the lowest BIC score suggests that I have between 17 and 20 clusters. I before running the analysis, I used the function recode.polyploids because of the two different ploidy levels.

Has anyone had a similar experience, or can suggest why the BIC score is disparate from the actual signal shown in the figure? I'd greatly listen to any advice folks can give on the matter.

-Jason Berg

U. of Maryland

[Zhian Kamvar]

Hi Jason,

Thank you for your question. It's important to point out that DAPC is not a function from poppr. It belongs to adegenet and it's best to use that forum for your DAPC query in the future: http://lists.r-forge.r-project.org/mailman/listinfo/adegenet-forum

The only function that you are using from poppr is recode_polyploids, which will remove zero-valued alleles that were used to facilitate import of polyploid samples.

Regarding your question about BIC values, perhaps this previous answer from the adegenet forum is helpful? http://lists.r-forge.r-project.org/pipermail/adegenet-forum/2011-June/000303.html

Best,

Zhian

--
You received this message because you are subscribed to the Google Groups "poppr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to poppr+un...@googlegroups.com.
To post to this group, send email to po...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/poppr/250e0540-8d04-4b44-8405-a4f1db4f2936%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.