[adegenet-forum] DAPC - visualising clusters when not separated geographically

[Ilaria Coscia]

Dear List members,
I have a question regarding the DAPC scatter.
I have 8 sampled populations (from 8 geographic locations), and the DAPC
analysis (as well as Structure), found 6 clusters. Now, these clusters
are not separate geographically, which means that each sampled location
has a proportion of each cluster. I wanted to see this on a scatter, to
see how different clusters were, hence I plotted the $ind.coord (I used
points()), and simply coloured the points by clustering this time (as
per DAPC analysis) and not by geographic location. I was expecting to
get same-colour points clustered close together (whereas in the first
DAPC, where colours were based on geog location, they were all
overlapping). Instead, colours remained all overlapping. I do not
understand how this explains the existence of 6 clusters, if they're not
there on the graph.
Maybe my assumption that re-colouring the points by cluster (and not
location) would present distinct groups was wrong?
I hope what I say makes sense...
I should say that I re-analysed from the start the same dataset too,
this time grouping individuals by cluster assignment and not by
geographical location, and I do get nicely separated clusters on the
scatter. I thought this was wrong in principle though, as I was basing
it on a new analysis.
Cheers
Ilaria
-- 
Dr Ilaria Coscia
Institute of Biological, Environmental and Rural Science (IBERS)
Edward Llwyd Building
Aberystwyth University
Aberystwyth
Ceredigion SY23 3DA
UK

tel +44 (0)1970 62 2298