I'm afraid I'm at a loss at discerning which element of the FactoMineR::PCA() output represents loadings and scores (sensu princomp, prcomp...). Any tips?

asked Jul 13 '10 at 14:14

Roman%20Lu%C5%A1trik's gravatar image

Roman Luštrik
1113


3 Answers:

The authors of the package have replied and inform me that:

scores are in res$ind$coord (individuals coordinates)
loadings are not given explicitely, you have to divide the variable coordinates by the square root of the eigenvalues

|sweep(res.pca$var$coord,2,sqrt(res.pca$eig[1:ncol(res.pca$var$coord),1]),FUN="/")|

FWIW...

answered Jul 15 '10 at 01:03

Roman%20Lu%C5%A1trik's gravatar image

Roman Luštrik
1113

As I read you, you are working on problems dealing with principal components analysis (PCA) and would like to know how to get the FactoMineR package to return loadings and scores like princomp does. Unfortunately, I do not know the answer, but the FactoMineR team seems to have some good references and tutorials, and if these do not help you can always contact someone in the FactoMineR team.

answered Jul 13 '10 at 15:15

John%20L%20Taylor's gravatar image

John L Taylor
61541518

edited Jul 13 '10 at 15:22

I swear I didn't see the FAQ section that talks about how to obtain scores and loadings (was it there a week ago?): http://factominer.free.fr/faq/index.html

(Jul 23 '10 at 16:16) Roman Luštrik

I didn't see it either. I think that they took your question as a prompt to update the FAQ. In any case, I'm glad it worked out.

(Jul 23 '10 at 17:28) John L Taylor

or if you really only want scores and loadings instead of rotations (factors), you could look into the pcaMethods package, which makes all of this trivially easy and implements NIPALS in C++ to boot (no more screwing around with imputation prior to doing an SVD). Just FYI.

answered Jul 14 '10 at 01:56

Tim's gravatar image

Tim
12

I'm comparing different PCA functions, so it would be desirable if I could get this to produce the output I'm interested in. I've searched the documentation but found nothing, so I've, per John's advice, contacted the author and see what he has to say. I will let you know if I get a reply.

(Jul 14 '10 at 02:25) Roman Luštrik

Tim, on a side note, isn't rotating a PCA (to get factors) in essence Factor Analysis?

(Jul 14 '10 at 02:26) Roman Luštrik

re: rotations: yes.

(Jul 14 '10 at 02:36) Tim
Your answer
toggle preview

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.