PCA of allele frequency

Hi,

Sorry if this is more a statistical question:

I'm running a PCA based on the correlation matrix on multiple allele  
frequencies (observations) measured in fish individuals from several  
locations of a large river (variables). Given this framework my  
questions are:

- Some alleles in different locations have zero frequency and the  
presence of many (double) zeros could induce an arching pattern. How  
many zeros are too many? Could this arch effect indicate the true  
ecological pattern? Is there any kind of transformation for these cases?

- I've used rda function and plot.cca for biplotting purposes. My  
interest is to make a statement on the variables (grouping,  
correlation among them etc), but when using scaling=2 (correlation  
biplot) variables are all close together and pointing in the same  
direction although some groups could be guessed. Using raw scores,  
i.e. scaling=0, for both observations and variables solves the  
visualization issue but, which is now the interpretation of the biplot?

I've attached the correlation biplot figure.

Many thanks in advance.

Jaime.


-------------- next part --------------