[adegenet-forum] Several questions

[Jombart, Thibaut]

Dear Marina, 

it is normal that the connection network is sometimes not plotted when plotting a spca object. As a matter of fact, connections are not displayed if there are more than 500 links: this avoid waiting for ages for the graphics to be drawn, especially since people may not be interested in displaying thousands of links of a connection network. As you say, it is moreover easy to plot the connection network in itself anyway. So, you are not mistaking, but it is not a bug.

Concerning your question about the Gtest/Ltest, I am glad you raised this issue. In fact, the entire dataset is tested for global/local structures when you perform these tests. It is not possible to use it to test the independent significance of different principal components. It tries to answer the question: "Are there any global/local structures in the data?", rather than "Is this principal component significantly spatially structured?".

Lastly, for the kind of distribution you described, I think you should consider aggregating genotypes from identical locations into 'populations' (using genind2genpop), and conduct the analysis at a spatial level. This is because even if you find a global structure, you won't know if it comes from similarity between individuals from the same population, or if it indicates spatial structures among populations. If you can't aggregate data by populations, then you may consider using transparency for the symbols, so that overlying symbols are not longer a big problem. You may want to use colorplot to do so: I've included transparency in the devel release of adegenet (use argument 'transp=TRUE', and alpha=[smthg between 0 and 1] to handle the level of transparency). You may also consider interpolation rather than plotting scores (see s.kde2d for an instance).

I hope this helps, do not hesitate if you have further questions.

Best,

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 mchiappero at efn.uncor.edu [mchiappero at efn.uncor.edu]
Sent: 19 January 2010 17:37
To: adegenet-forum at lists.r-forge.r-project.org
Subject: [adegenet-forum] Several questions

Dear Thibaut:

I found a seemingly minor problem when plotting the sPCA results:
everything works fine except the connection network graphic. It shows
the points representing my samples correctly, but no connections at
all. However, when I plot the connection network alone, it seems to be
OK. Am I making a mistake or is it a bug?
Also, I have two questions (I hope the last and I don't bother you any more):
First: how do I perform the Gtest and Ltest on the second principal
components?
I know this may seem dumb, and is probably a very basic "R thing",
but I can't realize how to do it and I couldn´t find any instruction
in the manual or tutorials.
The other is a request for advice: my dataset consists of individuals
(rodents) captured inside a city and in a nearby (~10 km)
agroecosystem. The problem is that in each of these two areas,
captures were conducted in several locations, and in each one,
individuals are very close together in relation to the total area.
Therefore the black or white squares in the sPCA graphic are highly
superimposed. Do you think it would be correct to indicate each
location with, say, a circle, and to separate each of the squares
representing the scores of each individual so that their size could be
appreciated?
Again, thank you very much
Marina

_________________________________________________________
Marina B. Chiappero
Genética de Poblaciones y Evolución
Facultad de Ciencias Exactas, Físicas y Naturales
Universidad Nacional de Córdoba
Av. Velez Sarsfield 299
(5000) Córdoba. Argentina
______________________________________________________________

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

_______________________________________________
adegenet-forum mailing list
adegenet-forum at lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/adegenet-forum