1
$\begingroup$

My aim is to construct a composite index for job quality from 33 indicators. I would like to set variable weights through Principal Component Analysis (PCA) and my data are longitudinal. My questions are:

  • Do you know how to manage PCA for panel data?
  • Can PCA be applied for binary and continuous data? only with binary data?
  • How can I define weights from PCA? I think correlation coefficients with the first Principal Component. But what about if I have more Principal Components?
$\endgroup$
1
$\begingroup$

In general, you want to apply the same weights to the 33 indicators at each time point. This is required to make scores comparable across time.

If you are using latent variable modelling approaches, then you can explicitly constrain common loadings to be the same at each time point.

If you want to do PCA, a simple approach would be to obtain the unstandardized weights at the first time point. Then apply these to every subsequent time period to derive your composite.

PCA can be used with binary and continuous data or a mix of the two.

$\endgroup$

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.