Help me understand the ouput of the posterior_epred function :-)

Interfaces brms
1

Hello Bayesian peoples :-)

I am running a dichotomous Bayesian Rasch model using the code below on a test consisting of 45 items & 225 respondents. My question is probably stupid but I am not able to understand properly the output of the posterior_epred function. I know I have 10199 columns in the outputs matrix, but what are they? Somebody have more information than those from the help files available in CRAN?
Thank you so much for your answer

# Run a Bayesian Rasch model
formula15 <- bf(Xb15 ~ 1 + (1 | It15) + (1 | P15))

prior15   <- prior("normal(0, 3)", class = "sd", group = "It15") +  prior("normal(0, 3)", class = "sd", group = "P15")

fit15    <- brm(formula = formula15,
                  data = X15,
                  family = brmsfamily("bernoulli","logit"),
                  prior = prior15)

# Posterior predictive check
ppe15 <- posterior_epred(fit15)
str(ppe15)
2

Sorry, it looks like your question fell through, did you manage to resolve your issue?

posterior_epred should return a matrix with “the number of rows in the dataset” and “number of samples” as dimensions, so what you describe is slightly weird…

3

I finally understood the output of this function. And you are right: I was unclear :-) Thanks for your answer