hdi() fails with brms model outputs about bayestestr HOT 9 CLOSED

That error seems weird - I cannot find any simliar call in any easystats package.

I also am unable to reproduce the example locally:

library(dplyr)
library(brms)
library(bayestestR)

df = data.frame(x = rnorm(1000), x2 = rnorm(1000)) %>%
  mutate(y = rnorm(1000) + x + x2 * .1)

model = brm(y ~ x + x2, data=df)

bayestestR::hdi(model)
#> Highest Density Interval 
#> 
#> Parameter   |       95% HDI
#> ---------------------------
#> (Intercept) | [-0.04, 0.09]
#> x           | [ 0.92, 1.05]
#> x2          | [ 0.07, 0.19]

from bayestestr.

Yeah, nothing of the error indicates that it could be an issue with the hdi() function.

from bayestestr.

Can you provide a reproducible example from your machine? (like, copy the code to clipboard and run reprex::reprex())

For me, it works fine:

library(brms)
library(bayestestR)
d <- data.frame(x = rnorm(1000), x2 = rnorm(1000))
d$y <- rnorm(1000) + d$x + d$x2 * 0.1
model = brm(y ~ x + x2, data = d)
bayestestR::hdi(model)
#> Highest Density Interval
#> 
#> Parameter   |       95% HDI
#> ---------------------------
#> (Intercept) | [-0.06, 0.06]
#> x           | [ 0.99, 1.12]
#> x2          | [ 0.02, 0.15]

^{Created on 2023-12-19 with reprex v2.0.2}

from bayestestr.

It appears that it is some kind of weird interaction with Jupyter. The code from @strengejacke example runs perfectly well in an R console or using Rscript but the same code gives the aforementoined error when run on the same machine within a Jupyter notebook cell or an R console window within Jupyter. I'll try to do some deeper comparisons to see if I can identify the source of the issue but interested in any thoughts.

from bayestestr.

@poldrack can you try running the following, just to try and locate the error:

library(brms)

d <- data.frame(x = rnorm(1000), x2 = rnorm(1000))
d$y <- rnorm(1000) + d$x + d$x2 * 0.1

H1 <- bayestestR::hdi(d)
print(H1)

model = brm(y ~ x + x2, data = d)

H2 <- bayestestR::hdi(model)
print(H2)

from bayestestr.

thanks, that code ran just fine, and helped isolate the problem: the error only occurs when bayestestR:hdi() is run within Jupyter without assigning the output to variable. Seems like this is a Jupyter problem rather than a problem with this function so I am fine closing this unless @strengejacke thinks there is something to be done here. thanks again for all your help!

from bayestestr.

Thank you for checking this more closely! I agree, it seems that nothing needs to be done from our side.

from bayestestr.

for the record I should note that my previous diagnosis was not exactly right: it turns out that it has to do with trying to print the outputs from Jupyter without surrounding the in an explicit print() statement. Thus, print(bayestestR::hdi(model)) works just fine as well.

from bayestestr.

without surrounding the in an explicit print() statement. Thus, print(bayestestR::hdi(model)) works just fine as well

Hi Russ, this is quite weird indeed. It seems like there could be a clash of methods where jupyter doesn't call bayestestR's print method automatically but something else, maybe some wrapper over print() to nicely output stuff but which unfortunately clashes (for some non-obvious reasons given the error trace)

from bayestestr.