[adegenet-forum] DAPC scatterplot vs membership probabilities

[Jombart, Thibaut]

Hi Carla,

sounds like an interesting problem.

Just to make sure, how did you derive the group membership probabilities? Manually, or did you use the output of predict.dapc?

If this is based on predict.dapc, we would need a reproducible example (small subset of data) to see what is going on. Assignment is based on the discriminant functions, so there should never be any mismatch, at least not with the first DF.

Cheers
Thibaut
________________________________
From: adegenet-forum-bounces at lists.r-forge.r-project.org [adegenet-forum-bounces at lists.r-forge.r-project.org] on behalf of Carla Riva Rossi [carla.rivarossi at gmail.com]
Sent: 10 December 2014 18:59
To: adegenet-forum at lists.r-forge.r-project.org
Subject: [adegenet-forum] DAPC scatterplot vs membership probabilities

I'm using DAPC to assess the origin of an introduced salmonid in Patagonia
    and I have problems interpreting my results.
    I used pre-defined groups corresponding to the native range source
    populations and then I applied the predict.dapc function to position the
    introduced individuals (that were not used in constructing the model) onto
    these discriminant functions (DFs). On the basis of the derived DFs, I
    derived for each introduced individual a membership probability to original
    source populations.
    Then, In the DAPC scatterplot the introduced individuals (population) fell
    more in the space with of native pop 1 than with pop 2, but the DAPC
    probabilites indicate more native pop2 ancestry. Which of these two results
    shoud I give more credit/support?
    Thanks a lot in advance!!!
    Carla.-

--
Dra. Carla Riva Rossi
CENPAT-CONICET
Blvd Brown 2915
Puerto Madryn (CP 9120)
Chubut - ARGENTINA
Tel: + 54-280 - 4883184  Interno: 1349
Fax:+ 54-280 - 4883543
email: rivarossi at cenpat-<mailto:rivarossi at cenpat.edu.ar>conicet.gob.ar<http://conicet.gob.ar>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/adegenet-forum/attachments/20141210/e2f23b36/attachment.html>