Principal Component Analysis: "Correlation Matrix is singular" error while conducting KMO

[SAKSHI MALIK]

Hi, Iam trying to create an index using variables capturing various economy related data. In order to reduce the data in groups. I conducted PCA When I conduct KMO after running PCA, the output just states the error "The correlation matrix is singular". Can I still go ahead with the results of PCA? Is KMO a necessary condition for PCA? If yes, how can I fix the singular matrix problem?
Can we go ahead with Principal Components Analysis (PCA) results if KMO result states that correlation matrix is singular?

[Nick Cox]

This is a bit like asking "Should I keep driving if one of the doors of my car falls off?".

The answer may be Yes if bad people are shooting at you too. We hope that's not happening. Otherwise No. Stop the car.

The correlation matrix should show you what is wrong. So, fix the choice of variables, then revisit the PCA.

[SAKSHI MALIK]

I get it now, Nick Cox. Followed your suggestion, and based on the results of the correlation matrix, dropped some variables with high correlation. Yet, KMO is displaying the same error. Here is the correlation matrix for the new set of variables.Which variables according to you sir, if deleted, might help in resolving the problem of "singular correlation matrix" in KMO?

[Nick Cox]

Sorry, I really can't advise except to comment that the decision should be substantive as much as statistical. Even someone in your unstated field couldn't judge from these variable names what is going on.

Just skimming through these correlations I see lots of weak correlations. PCA is not much better than anything else at identifying pattern where is a lot of noise.

[SAKSHI MALIK]

Well yeah, that's my bad. I am new at statalist, will take my time to understand how to post.