Interaction Effects Reference Coding

Hi there,

I read the summary on fitlme (https://de.mathworks.com/help/stats/fitlme.html). I am having a question regarding the reference coding.

In the example "Longitudinal Study with a covariate" the effect of Prorgam B is significant indicating a stronger effect of B than A. I want to confirm that my understanding of interaction effects is correct. If for example Program B: Week would have been significant, that would mean that the effects of weeks is different under Programm B compared to A (and not compared to all other programs?)

Thank you!

Answer 1

[praguna manvi]

Hi @Laurie König,

In the example "Longitudinal Study with a covariate", the model uses Program A as the reference category, this means that the coefficients for Programs B, C, and D are compared to Program A.

The coefficient for "Program_B" is significant (p-value = 0.0070394), indicating that Program B has a significantly different effect on weight loss compared to Program A, averaged over all weeks.

The interaction term "Program_B:Week" is not significant (p-value = 0.68521), meaning there is no significant difference in the effect of "Week" on weight loss between Program B and Program A. Each interaction term specifically compares the effect of "Week" for that program to Program A, the reference category.

Hope this helps!