How to plot PCA output?

I don't know the answer, Jessica gave some insight.

I avoid the biplot at all costs, because IMHO it violates one of the
tenets of good graphic design:  It has two entirely different scales on
axes.  These are maximally confusing to the end-user.  So I never use it.

If it is gene expression data, have you looked in Bioconductor for
something that will help you?  Maybe runPCA in package EMA?

Bryan

On May 7, 2012, at 9:57 AM, Christian Cole wrote:

Hi Bryan,


Many thanks for the replies.

The data is gene expression data for 36 samples over 11k genes.

I see that I can plot PC1 vs PC2 by using $x, but compared to biplot() I
can see that the range of values are different. For example, if I use
plot() the PC1 scale ranges from -150 to 150 whereas in biplot() it
scales
from -0.4 to 0.4. Do you know what scaling biplot() uses? Does it even
matter?
Cheers,

Chris


On 07/05/2012 14:36, "Bryan Hanson" <hanson at depauw.edu> wrote:

Christian, is that 36 samples x 11K variables?  Sounds like it.  Is
this
spectroscopic data?

In any case, the scores are in the list element $x as follows:

answer <- prcomp(your matrix)

answer$x contains the scores, so if you want to plot the 1st 2 pcs, you
could do

plot(answer$x[,1], answer$x[,2])

Because the columns of answer$x contain the scores of the PCs in order.

[I see Jessica just answered...]

If you want the loading plot, it's going to be interesting with all
those
variables, but this will do it:

plot(1:11000, answer$rotation[,1], type = "l") # for the loadings of
the
1st PC

Depending upon what kind of data this is, the 1:11000 could be replaced
by something more sensible.  If it is spectroscopic data, then replace
it
with your frequency values.

By the way, plot(answer) will give you the scree plot to determine how
many PCs are worthy.

Good luck. Bryan

***********
Bryan Hanson
Professor of Chemistry & Biochemistry
DePauw University

On May 7, 2012, at 6:22 AM, Christian Cole wrote:

I have a decent sized matrix (36 x 11,000) that I have preformed a PCA
on
with prcomp(), but due to the large number of variables I can't plot
the
result with biplot(). How else can I plot the PCA output?

I tried posting this before, but got no responses so I'm trying again.
Surely this is a common problem, but  I can't find a solution with
google?


The University of Dundee is a registered Scottish Charity, No:
SC015096

The University of Dundee is a registered Scottish Charity, No: SC015096