Why does hypothesis return different CIs than summary?

I would expect that for simple hypothesis of the form model_parameter < 0 hypothesis would return CIs mirroring the model summary. However, there tends to be a small difference… am I misunderstanding something about how hypothesis or summary works?

An example:

library(tidyverse)
library(brms)
fit <- brm(y ~ x, data = data.frame(x = rnorm(30)) %>% mutate(y = rnorm(30, x)))
summary(fit)
# Population-Level Effects: 
#           Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
# Intercept    -0.13      0.22    -0.55     0.30 1.00     3647     2503
# x             0.91      0.21     0.49     1.33 1.00     3593     2549

hypothesis(fit, "x < 0")
#   Hypothesis Estimate Est.Error CI.Lower CI.Upper Evid.Ratio Post.Prob Star
# 1    (x) < 0     0.91      0.21     0.56     1.26          0         0    

The two sided hypothesis x = 0 will give the same CIs. The CI in your hypothesis example is a one sided CI.

2 Likes