Brms Distributional Models vignette example giving weird results

#1

Hi,

I’m trying to run the code in the vignette: “Estimating Distributional Models with brms”
(2019-03-14, https://cran.r-project.org/web/packages/brms/vignettes/brms_distreg.html). I changed the numbers a bit for higher resolution/precision:

group <- rep(c(“treat”, “placebo”), each = 1000)
symptom_post <- c(rnorm(1000, mean = 10, sd = 3), rnorm(1000, mean = 0, sd = 1))
dat1 <- data.frame(group, symptom_post)
head(dat1)
fit1 <- brm(bf(symptom_post ~ group, sigma ~ group), data = dat1, family = gaussian())
summary(fit1)

This gives:
Population-Level Effects:
Estimate Est.Error l-95% CI u-95% CI Eff.Sample Rhat
Intercept 0.06 0.03 -0.00 0.12 5689 1.00
sigma_Intercept 0.01 0.02 -0.04 0.05 3934 1.00
grouptreat 9.92 0.10 9.72 10.11 2610 1.00
sigma_grouptreat 1.11 0.03 1.04 1.17 3911 1.00

The effect sizes look good, but the sigmas should be 3 and 1, not 1 and 0, right? Why is this happening?

Thank you,

Hershel

  • Operating System: Mac OS 10.13.6
  • brms Version: brms_2.8.0
0 Likes

#2

sigma is modeled on the log scale, which you also see at the start of the summary output.

0 Likes

#3

Oops, thank you!

-Hershel

0 Likes