0

I perform principal component analysis (PCA) on a dataset with 120 variables.

If I want to know if two variables A and B from my original dataset are correlated, how could I answer this question after performing PCA? As I understand from the answers to my previous question (Does each eigenvalue in PCA correspond to one particular original variable?), PCA transforms my variables, and original A and B are not available anymore.

Community
  • 1
Linda
  • 31
  • 1
  • 2
  • I edited your question, please check if it corresponds to what you wanted to ask. Also, if you are satisfied by one of the answers to your previous question, consider "accepting" it by clicking on a green tick to the left of the answer. – amoeba Jul 17 '15 at 11:05
  • 1
    In many cases the actual *input* to PCA consists of the correlation matrix. That would render this question moot. In order for it to be understandable, then, you will need to edit it to indicate precisely what values you have available in the PCA "results" as well as what information you have input to it (and whether that input information is available). – whuber Jul 17 '15 at 13:28

1 Answers1

1

The variables in your data set are correlated the way they are correlated, regardless of whether you do PCA or not. If you want to see if A and B are correlated, then you ask your software to print the correlation for those two variables.

The goal of PCA is dimension reduction. The components are each linear combinations of the variables. If you want the original variables, use them.

Peter Flom
  • 94,055
  • 35
  • 143
  • 276