I’m attempting to calculate the 89% credible interval for a posterior distribution of an estimate of a proportion (and so the result is necessarily bounded by [0,1]) in R using the bayestestR package.

However when using the HDI() from the bayestestR library the result is CI low 5.15e-260 & CI high 1.09 (faint blue lines on the plot below).

ci_hdi <- distributionDF%>%
  filter(variable=="posterior")%>%
  select("Mechanism")%>%
  hdi(ci = 0.89, method="HDI")

(There are 1560 observations for this variable, so I have included the data here but see below for an MRE)

Prior, Likelihood and Posterior Distributions with CI

I assumed I was making an error somewhere so I reproduced the example from the vignette, and got a similar result:

library(bayestestR)
library(ggplot2)
ci_hdi <-ci(posterior, method = "HDI")
ci_eti <-ci(posterior, method = "ETI")

posterior <- distribution_beta(1000, 6, 2)

output <- estimate_density(posterior, extend = TRUE)

ggplot(
  output, aes(x=x,y=y))+
    geom_area(fill = "yellow")+
    theme_classic()+
    #HDI in blue
    geom_vline(xintercept =ci_hdi$CI_low, colour = "royalblue", linewidth = 3)+
    geom_vline(xintercept =ci_hdi$CI_high, colour = "royalblue", linewidth = 3)+
    #ETI in red
    geom_vline(xintercept =ci_eti$CI_low, colour = "red", linewidth = 3)+
    geom_vline(xintercept =ci_eti$CI_high, colour = "red", linewidth = 3)

Resulting posterior distribution with CI

Is there occurring because the densities (variable) is not normalised to one?