[adegenet-forum] Use of Lag spca$ls values.

[Jombart, Thibaut]

Dear Nevil, 

The lagged principal component ($ls) is a smoothed version of the principal components ($li), obtained by replacing the score at one site by the mean score of its neighbouring sites. When the pattern is clear, $ls will match perfectly $li. When the pattern is more noisy, $ls will help recognising the underlying pattern. Technically, if 'v' is a principal component of sPCA, and M a matrix of neighbouring weights, then the lagged score v' is computed as:
v' = Mv

Best regards

Thibaut
________________________________________
From: Nevil Amos [nevil.amos at gmail.com]
Sent: 07 September 2010 12:00
To: adegenet-forum at lists.r-forge.r-project.org
Subject: Use of Lag spca$ls values.

  It is noted in the draft manual that the "The lag vectors of the
scores can be displayed graphically instead of basic scores
so as to better perceive global structures".  In some cases I have
examined the pattens showed by the lag vectors are similar to those
shown by the spca$li score.

In some other cases though the pattern is quite different.  How should I
interpret this?



If the score represents the maximised  spatial spatial
autocorrelation*genetic variance, what does the lag Vector represent.

I have read Jombart et al 2008, but it is not clear to me, probably as a
result of my limited mathematics!

cheers

Nevil Amos