Does running a no-intercept model with multiple categorical predictors and interaction between the predictors result in valid estimates?
I am modeling differences within paired data and am using a no-intercept linear model. I have a continuous predictor (A) and a categorical predictor with three levels (B). I am also interested in interaction between the predictors and a two-level categorical variable (Z).
When I run a model with interaction between (A and Z) and (B and Z), the results should be equivalent to the models stratified by Z (since there are no other variables in the model). The results are identical for the reference level of the interaction variable (Z). For the other level of of the interaction variable (Z), the results are identical for the continuous variable (A), but not for the levels of the categorical variable (B).
I believe the results of the stratified models are correct and the difference may be due to using a no-intercept model. Is it possible to use/interpret a no-intercept model with multiple categorical predictors and interaction between the categorical predictors?