Prior predictive check for the effect of a categorical predictor variable

Interfaces brms
1

Hi everyone,

I’m running a hierachical logistic regression with one categorical predictor:

library(brms)

dat$structure <- factor(dat$structure, levels = c("simple", "complex"))

prior = c(prior(normal(2.64,0.43), class = "b", coef = "Intercept"), 
              prior(normal(-1.14,0.49), class = "b", coef = "structurecomplex")
              )

model <- bf(supplied ~ 0 + Intercept + structure + (1 + structure | ID) + (1 | item),
  family = bernoulli)

fit <- brm(
  formula = model,
  prior = prior,
  data = dat,
  chains = 4,
  cores = 4,
  warmup = 1000,
  iter = 4000,
  control = list(adapt_delta = 0.99)
)

Here is the data set:

dat.csv (33.9 KB)

I’ve done a prior predictive check on the intercept:

model_PPC <- brm(
  formula = model,
  prior = prior,
  data = dat,
  sample_prior = "only",
  seed = 2780
)

This should be OK. If not, please let me know.

What I need help with is this: what would be the code for doing a prior predictive check on the effect of the predictor variable?

I’d be very grateful for any assistance with this.

  • Operating System: Windows 11
  • brms Version: 2.16.1