1
$\begingroup$

I performed a PCA on credit default swaps spreads and I want to predict the first principal component scores. My two doubts are:

  • If I use the covariance matrix in my PCA I'll have a predict first principal component that is measured in the units of the original variable.

  • If I use the correlation matrix I'll have a predict first principal component score that is in the standardized form.

Is this correct?

$\endgroup$
  • 3
    $\begingroup$ It may help you to read: PCA on correlation or covariance? $\endgroup$ – gung - Reinstate Monica Oct 28 '15 at 23:02
  • 1
    $\begingroup$ Hi @Ana, welcome to CrossValidated. +1. If you perform PCA on covariance and if all your variables are in the same units, then yes, you can say that principal component scores are in the same units. If you perform PCA on correlation, then essentially you have made all your variables dimensionless (unitless) by standardizing them. The PC scores are not going to be standardized (i.e. they will not have unit variance!), but they will also be unitless. $\endgroup$ – amoeba says Reinstate Monica Oct 29 '15 at 1:04

Your Answer

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

Browse other questions tagged or ask your own question.