Significant treatment*pretest interaction, but interaction not shown with regression lines


I have found that the interaction treatment by pretest is significant. However, the interaction is not depicted on the scatterplot (see attachments). Am I doing something wrong or can I say that the interaction is outside the range of possible test scores (0-10) and that the assumption of homogeneity of regression slopes is met?

My next step would be to calculate the cross over point and to determine the regions of significance. When the crossover point is outside the range of data, does that mean that the two groups are not different across the range (even though there is an interaction)?



Last edited:


Less is more. Stay pure. Stay poor.
When examining an interaction term it is standard practice to keep the main effects terms in the model. What happens when you add them back in?