I am new to
brms, trying to figure out its behaviour in details. I was trying to run essentially an anova-like model with random effects (anova-like in the sense that all IVs are factors). For example,
brm(DV ~ A * C + B * C + (1|item) + (1|participant), ...
hypothesis() are truly amazing analytic tools. However, when you deal with factors and, in particular, their interactions, that’s a bugger… It’s really hard to help students wrap their head around it. In addition to that, I am a bit stubborn and I really despise when people, for practical purposes, re-level their factors to figure out what contrasts are significant. There is no need for that. In frequentist framework, one could make use of Wald’s test etc. Here, I sense, the two functions can be of much use, but still one need to be careful around the intercept. All other levels are relative change to that reference.
If one have a simple dummy variable,
0 + Intercept + Dummy + ... is of help (another nice one!). But when you have factors with more levels and their interactions, things are far from straightforward.
What would be the brms way of making this less painful? Running another model without intercept and then use
hypothesis()? Something else?
- Operating System: macOS
- brms Version: 2.14.4