I conducted a mixed model ANOVA analysis with repeatedly measured data. Should Include all possible combination between variables? Or include interactions of interest based on research hypothesis?
There are two ways of looking at this issue. One is that you run all combinations of independent variables including polynomials and interactions, and find the best fitting model. The second approach would be to choose variables (and interactions and polynomials) based on your content expertise. In either case, you should provide rational for why you chose the given variables, and you should report them in full, if they were included in the model. The simple reason is because your results should be transparent reproducible...
I think you should only put in the interactions that are of interest based on your hypothesis, since that what you're looking for. The more comparisons you make, the higher the chance of making a type 1 error.
I am by no means a statistician, so hopefully a true statistician replies as well.
There are two ways of looking at this issue. One is that you run all combinations of independent variables including polynomials and interactions, and find the best fitting model. The second approach would be to choose variables (and interactions and polynomials) based on your content expertise. In either case, you should provide rational for why you chose the given variables, and you should report them in full, if they were included in the model. The simple reason is because your results should be transparent reproducible...
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA