[R-sig-ME] Simulating beta-distributed outcome in glmmTMB

[Simon Harmel]

Hello all,

I'm trying to simulate about 50 datasets each with a beta-distributed
outcome (Y) according to the R code below.

For each dataset, I would like to randomly determine the parameter
values for "beta=" and/or "theta=" arguments in my
glmmTMB::simulate_new() call such that I avoid getting Y values that
are close to 0 or 1.

I was wondering how you all suggest I can achieve this?

Thanks, Simon

## R code for 1 dataset:
library(tidyverse)
library(glmmTMB)

n_participant=100
n_time=3
groups = c("A","B", "Control")

set.seed(123)

dd <- data.frame(ID = ID <- rep(1:n_participant, each = n_time),
                 time = factor(rep(0:(n_time-1), n_participant),
levels=0:(n_time-1)),
                 group = factor(sample(groups, n_participant, replace=TRUE)[ID],
                                   levels = groups))
dd$Y <- simulate_new(
  ~ group*time + (time | ID),
  family = beta_family(),
  newdata = dd,
  newparams = list(beta = c(.2, .2, .3, .5, .1, 1.7, .9, .3, .1),
                   theta = c(-1.1, -1.25, -1.4, -1.2, -1.4, -1.3)))[[1]]


model <- glmmTMB(Y ~ group*time + (time | ID),
       family = beta_family(),
        data = dd)

VarCorr(model)