PCA is normally not well-suited for non-metric data. The reason is that PCA relies on covariances or correlations, which require numeric, continuous variables. The basic assumptions of PCA (such as the calculation of Euclidean distances) do not hold for non-metric data.

You could either try to transform your data somehow so that you end up with numeric data (and thus you could calculate those distances). Alternatively try Multiple Correspondence Analysis (MCA) or Factor Analysis of Mixed Data (FAMD)!

Hey Neelam,

Principal components analysis (PCA) is usually applied to metric data—like interval or ratio data—where the distances between values matter. This is because PCA looks at variance and assumes that the data has meaningful numerical relationships. If you apply PCA to non-metric data (like categorical data), it won't give useful results, since those kinds of data don’t have measurable distances between categories.

For non-metric data, you'd use techniques like Multiple Correspondence Analysis (MCA) instead, which is designed for categorical data.

Hope this clears it up!

@Igor Faddeenkov & Anitha Roshan Sagarkar for your valueable responses.