Re: [R] prcomp - surprising structure

On Oct 3, 2013, at 16:30 , Hermann Norpois wrote: Thanks for answering. I already started hunting. But my first doubt was if I used prcomp correctly (and this is in the moment my most important point). So far as I understood your answer is yes. Is that correct? Yes. There are a couple

[R] prcomp - surprising structure

Hello, I did a pca with over 20 snps for 340 observations (ids). If I plot the eigenvectors (called rotation in prcomp) 2,3 and 4 (e.g. plot (rotation[,2]) I see a strange column in my data (see attachment). I suggest it is an artefact (but of what?). Suggestion: I used prcomp this way:

Re: [R] prcomp - surprising structure

It's not so obvious to me that this is an artifact. What prcomp() says is that some of the eigenvectors have a lot of activity in some relatively narrow ranges of SNPs (on the same chromosome, perhaps?). If something artificial is going on, I could imagine effects not so much of centering

Re: [R] prcomp - surprising structure

Thanks for answering. I already started hunting. But my first doubt was if I used prcomp correctly (and this is in the moment my most important point). So far as I understood your answer is yes. Is that correct? I am puzzled by the fact that these columns are more or less in the middle of my