1
$\begingroup$

My ultimate goal is to run a cluster analysis on a data set with > 1 million records. The input variables for the cluster analysis will be the results of a Principal Component Analysis, as well as other variables not included in the PCA, for a total of maybe 10 variables input into the clustering (the variables I input into the PCA were all very highly correlated with one another while the other variables are not so I chose not to include them in the PCA).

#read data
mydata <- read.csv('mydata.csv') 

#import library for robust methods because my data contained outliers
library(rrcov) 

#run robust PCA method called PcaCov
pcaR <- PcaCov(~., mydata, na.action=na.omit, center=TRUE, scale = TRUE, k=8)

#look at results
summary(pcaR)
screeplot(pcaR)
pcaR@loadings

From the results, I have decided I would like to retain the first three components, which capture ~87% of the total variance in the dataset.

Now I want to extract/save/export these first three components for use in the cluster analysis with my other variables. How do I do this?

$\endgroup$
1
$\begingroup$

For each variable obtained by PCA you have a loading vector (for example $v=(1,-2,5,5)$ this vector define your new variable as combination of the original ones. $x_1-2x_2+5x_3+5x_4$. You can define a new matrix where the variables are obtained as the linear combination defined by the loadings obtained with PCA. So for example $z_1=x_1-2x_2+5x_3+5x_4$.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks @Donbeo. I have edited the code above based on your suggestion (basically if I understand correctly, I multiply the loadings by the original variables?). I don't do anything with the scores?? $\endgroup$ – sth May 24 '14 at 19:12
  • $\begingroup$ No wait this is not what I meant. If the function allows you to extract the scores than this is already the representation of X in the new base. This would probably be of the same number of columns of the original X. Be careful that usually PC and loadings are the same thing. Be also careful that probably you are not doing the matrix multiplication with the right command $\endgroup$ – Donbeo May 24 '14 at 19:16
  • $\begingroup$ Ok so I guess I just export the scores! $\endgroup$ – sth May 24 '14 at 19:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.