Which type of ANOVA should I select for a Power Analysis of a Factorial Design

Question

I am trying to develop a function to simulate datasets, in order to perform power analysis on a Factorial Design (for example 3x5 design). In my case, I'll be interested in testing both differences in the main effects (A and B), as well as for the interaction (A*B). I will also work under the assumption of a balanced design across all 15 groups.

In this scenario, which type of ANOVA should be the one to use? I have read several sites on Type I vs Type II vs Type III but, mostly state such difference is important mostly on unbalanced designs, with some places stating "on a balanced design, they will give the same results", which does not seem to be the case.

Thanks in advance

Answer 1

I reviewed the simulations I was running and understood that I was running the ANOVA type III wrongly (using car::Anova function).

In a balanced design as the one I was performing, ANOVA I, II and III should yield the same results.