1

I am running KMO (a function of the "psych" package that calculates the Kaiser, Meyer, Olkin Measure of Sampling Adequacy) in R with 75 variable 7232 observation datasets. In two of 16 cases I get this error:

Error in solve.default(r) : Lapack routine dgesv: system is exactly singular: U[74,74] = 0 matrix is not invertible, image not found

Nevertheless, the function gives a solution.

Should I worry about the error although I get a solution? If yes, what kind of error can be causing the error and how can I solve the problem?

Thank you very much.

EDIT: The error is due to the matrix's rank being 72, instead of 75 (n). Would it be wise to scale the values of the observations so that this error disappears, even if all the variables are indicator variables (in order to perform prcomp I have already centerred them)?

In this link you can find and Excel document containing one of the problematic matrices:

https://drive.google.com/open?id=1x0a4eo0V_ujR0tyNoDcLDkLQnLO0t4P9

Maite CD
  • 21
  • 5
  • Maite, consider making this problem reproducible. To do so, you'll need to give us sample data and the minimum code required (given your data) to reproduce your error. It really helps if you can reproduce it with *small data*, though it may be hard since it appears this will be specific to its un-invertibility. Refs: https://stackoverflow.com/questions/5963269/, https://stackoverflow.com/help/mcve, and https://stackoverflow.com/tags/r/info. – r2evans Aug 28 '18 at 15:58

0 Answers0