3 categorical variables analysis

Question

apologies if this question is obvious. I am trying to determine what type of analysis I should run for an observational study I am doing. The variables are sleep (sleep deprivation <7 hours or adequate sleep 7-9 hours), physical activity levels (level 1-4, increasing levels is more activity), and diabetes prevalence (this is the dependent variable).

My group and I are wondering if a chi-square test with a 3x3 contingency table would be more appropriate, or a logistic regression. The point of this study is to see if the independent variables: sleep and physical activity levels, have an effect on T2D prevalence in people aged 65+ and if so which one has more of an effect. Thank you!

Answer 1

This is not a 3 by 3 contingency table. It would become a 3-dimensional array. But that's a minor point.

The answer really depends on what you would like to achieve, but at any rate, I would recommend that you go with logistic regression. It provides a nifty way of interpolating predictor values and it returns more information than just performing a test. Furthermore, if you have predictors as continuous variables before dichotomizing sleep (I assume physical activity was formulated as a question with a choice of 4 levels from the get-go), then it is beneficial to use the continuous variable, as 7 hours may not be a sensible cutoff point.