1. Interactions introduce collinearity into a multiple regression and should be removed from the model if not statistically significant.
2. If neither the interaction nor the dummy variable is statistically significant in an analysis of covariance, then there’s no lurking factor that confounds the results of the related two-sample t-test.
3. To be a confounding variable, the variable must be related to Y and to the dummy variable that indicates group membership.
4. A major assumption of the use of regression with dummy variables is that the size of the two groups must be approximately the same in order to increase the variation of the dummy variable.
5. To check the similar variances condition in models with a dummy variable, use comparison boxplots of the response Y versus the categorical variable.
6. To fit a multiple regression that compares the mean values of five groups identified by a categorical variable requires using five dummy variables.
7. An analysis of covariance involving four groups requires that the residuals associated with each group have similar variances.