Yes. However, the expected count should be > 5.

Hi, using simple regression based on crosstab data is not recommended but using generalized linear model is recommended. Would you write us more about your data?

Hello Getasew,

I agree with Jochen Wilhelm's response, especially if you're contemplating a nominal x nominal type of analysis. However, do remember that, if you have a nominal variable with only two levels, and one has no cases, then that characteristic is a constant for your data set, and not a variable (and therefore should be reported as such, but omitted from model-building exercises).

Good luck with your work.