I am trying to use principal component analysis (PCA) to reduce dimensionality before applying linear regression. The problem is that my first 10 components are so weak (explaining only tiny variances - the 10th component's cumulative is 0.2577). What else could I try to apply the PCA?

Also I understand the whole process is referred to as principal component regression (PCR). Is there a tutorial or example I could learn in Stata/R?


migrated from stackoverflow.com Jun 22 '15 at 8:05

This question came from our site for professional and enthusiast programmers.

  • 2
    $\begingroup$ Do you need to use PCR, or is the question about what else you can try instead? $\endgroup$ – amoeba Jun 22 '15 at 8:23
  • 1
    $\begingroup$ First point: how many predictors do you have in total? If you have 1000 variables, 25% of variance explained by the first 10 is not that bad. Second point: PCR is just a regular regression where the covariates are the PCs, so you should only read about regression. $\endgroup$ – Antoine Jun 22 '15 at 9:36
  • $\begingroup$ I have to use PCA as the next step is to do the PCR. I guess that I was a bit lost when I saw the first 10 components were so weak and only explained so little variance. $\endgroup$ – Jun No Jun 22 '15 at 23:41
  • $\begingroup$ Antoine, thank you for the suggestion. I would try and see if I could go further and complete the PCR. $\endgroup$ – Jun No Jun 22 '15 at 23:43

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.