I’m new to this forum, so I hope I don’t come across as an idiot. :). This seems like such a basic question, but I’m just not confident since I’m new to multinomial models.

It’s a simple situation - one two-level predictor being used to predict which of four categories will be chosen (using brms, family=categorical). I know how to test for a specific response using emmeans and dpar, but is there a way to determine if the predictor had an effect on the *distribution* of response probabilities (a type of omnibus test) before diving into specific comparisons for each response category?

Here’s some example output

```
Group-Level Effects:
~Subject (Number of levels: 40)
Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
sd(mu2_Intercept) 1.23 0.15 0.97 1.56 1.00 1524 1944
sd(mu3_Intercept) 0.79 0.10 0.63 1.01 1.00 1728 2334
sd(mu4_Intercept) 1.16 0.14 0.92 1.46 1.00 1521 1893
Population-Level Effects:
Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
mu2_Intercept 1.23 0.20 0.86 1.63 1.00 1531 1903
mu3_Intercept -0.25 0.13 -0.50 0.01 1.01 1816 1980
mu4_Intercept -0.06 0.19 -0.45 0.31 1.00 1721 1832
mu2_Injury1 1.29 0.19 0.91 1.67 1.00 1494 1795
mu3_Injury1 0.21 0.13 -0.04 0.46 1.00 2073 1687
mu4_Injury1 0.34 0.19 -0.02 0.70 1.00 1610 1979
```

I thought the solution might be in using the hypothesis command:

hypothesis(modelname, “mu2_Injury1 + mu3_Injury1 + mu4_Injury1 = 0”)

But, I’m not confident that this is doing what I want. Any thoughts?

I have considered the option of just running a model without my predictor and comparing the two models, but I have 16,000 simulated data sets I’ve already analyzed using brms and am not looking forward to running 16,000 more!

I’m running this on a Mac with brms version 2.15.0