# CWT coefficients as features for ML algorithms

I use CWT coefficients as features in ML algorithms and then I did the feature selection using the chi-square test but recently I figured out that the chi-square test can only be applied for categorical variables. But, can we assume the CWT matrix values as categorical features and then apply the test?

• Hi and Welcome do DSP.SE, although your question has an aspect of signal processing, however the nature of your question is statistically. Nevertheless, in my opinion CWT coefficients are not categorical as the numbers/differences/intervals are comparable to each other and additionally these coefficients are continuous. – Irreducible Dec 3 '19 at 5:49
• Yes, I figure out unless if we binned the coefficients this assumption is invalid due it's a violation of the concept of independence which governs this test. – hasi Dec 6 '19 at 17:47