reporting of interactions and main effects

#1
Hi

I have a general question. When you include interaction terms in a regression model together with main effects and the interaction terms are significant, do you only report on the significant interaction effect? Do you need to report or mention whether the dependent variables involved in the interaction term are significant or not?

Let's say I have dependent variables a and b and the interaction term a*b and all are included in the model. The results are that:

a*b is significant
a is not significant
b is not significant overall - but b is the level of education and although the overall p value for education is not significant, the p value for one of the education levels is significant (bachelor degree compared to the reference group; high school).

In this case, do I only report on the results of the interaction term a*b. Do I also need to report on the results for b because one of the categories of b is significant even when the interaction term is included.

I am very confused and would appreciate your help.
 

bugman

Super Moderator
#2
Because a significant interaction term is indicating that the level of one factor is dependant on the level of another, then reporting manin effects makes little sense.

Report the interaction term but include the ANOVA results in a table with appropriate graphs and your reader will be very happy.