I have managed to get an ordbetareg model to run and am trying to use the “hypothesis” command. My model has two numeric predictors (Dose and Day) and one two-value factor predictor (Genotype). Hypothesis works like a charm for the numeric predictors, to wit: hypothesis(Thigmo.mod, “Dose < 0”) gives an output. However, when I try hypothesis(Thigmo.mod, “Genotype < 0”), I get “Error: Some parameters cannot be found in the model: ‘b_Genotype’” What syntax should I use?
Also, in this case, I have good theoretical reasons to test one-way hypotheses. The help for “hypothesis” is utterly unhelpful on how to do a hypothesis test that a variable’s effect is either greater than or less than, but not zero.
The call to hypothesis needs to reference the names brm uses internally for the variables (ordbetareg uses the brm function from the brms package to fit the model). You can see the variable names by running variables(my.model) (where my.model is the model fit object) or by just typing my.model in the console, which will return the model summary, which includes the names for each parameter.
In the case of a factor variable, the individual levels (excluding the reference level) are each converted to dummy variables. For example, if Genotype has two levels A and B, where A is the reference level, then the parameter in the model will be GenotypeB and you could run hypothesis(my.model, "GenotypeB < 0"). (If the intercept is excluded from the model, then GenotypeA will also be an explicit parameter in the model.)