A researcher studies the association between students’ rating of professors (good, fair, poor) and class size (small, large). If all the respondents come from the same handful of classes, what...

A researcher studies the association between students’ rating of professors (good, fair, poor) and class size (small, large). If all the respondents come from the same handful of classes, what underlying assumption of the χ²
test would be violated?

Since regression methods require the errors to be normally distributed, you apply the χ²
goodness of fit test for normality to the residuals, and obtain a p value of 0.001. What does this suggest should be done?