Error when using rstanarm::product_normal

Modeling
1

I’m interested in using the rstanarm::product_normal()prior, but it error.
What am I doing wrong here?

library(rstanarm)

m <- stan_glm(mpg ~ am,
              data = mtcars,
              prior = product_normal(2, 0, 1),
              family = gaussian())
#> Error in new_CppObject_xp(fields$.module, fields$.pointer, ...) : 
#>   Exception: mismatch in number dimensions declared and found in context; processing stage=data initialization; variable name=num_normals; dims declared=(1); dims found=()  (in '/data/hyperparameters.stan' at line 18; included from 'model_continuous' at line 56)
#> failed to create the sampler; sampling not done
#> Error in check_stanfit(stanfit) : 
#>   Invalid stanfit object produced please report bug
2

Sorry for taking so long for an answer to pop up! I think that the problem is the 2 in your call to product_normal()? According to the documentation,

Each element of df must be an integer of at least 2 because these “degrees of freedom” are interpreted as the number of normal variates being multiplied and then shifted by location to yield the regression coefficient. Higher degrees of freedom produce a sharper spike at location .

Does the above imply that you should provide a vector, i.e., degfree <- c(2,2) or something like that?

3

@torkar Nope… Still getting an error…

library(rstanarm)

m <- stan_glm(mpg ~ am,
              data = mtcars,
              prior = product_normal(c(3,3), 0, 1),
              family = gaussian())
#> Error in new_CppObject_xp(fields$.module, fields$.pointer, ...) : 
#>   Exception: mismatch in dimension declared and found in context; processing stage=data initialization; variable name=prior_df; position=0; dims declared=(1); dims found=(2)  (in '/data/hyperparameters.stan' at line 10; included from 'model_continuous' at line 56)
#> failed to create the sampler; sampling not done
#> Error in check_stanfit(stanfit) : 
#>   Invalid stanfit object produced please report bug

Created on 2021-02-01 by the reprex package (v0.3.0)

4

Hmm, it seems that scale also needs to be a vector?

5

It seems like all options give some error (slightly different Exceptions…):

library(rstanarm)

m <- stan_glm(mpg ~ am,
              data = mtcars,
              prior = product_normal(c(3,3), 0, c(1,1)),
              family = gaussian())
#> Error in new_CppObject_xp(fields$.module, fields$.pointer, ...) : 
#>   Exception: mismatch in dimension declared and found in context; processing stage=data initialization; variable name=prior_scale; position=0; dims declared=(1); dims found=(2)  (in '/data/hyperparameters.stan' at line 2; included from 'model_continuous' at line 56)
#> failed to create the sampler; sampling not done
#> Error in check_stanfit(stanfit) : 
#>   Invalid stanfit object produced please report bug

m <- stan_glm(mpg ~ am,
              data = mtcars,
              prior = product_normal(c(3,3), c(0,0), c(1,1)),
              family = gaussian())
#> Error in new_CppObject_xp(fields$.module, fields$.pointer, ...) : 
#>   Exception: mismatch in dimension declared and found in context; processing stage=data initialization; variable name=prior_scale; position=0; dims declared=(1); dims found=(2)  (in '/data/hyperparameters.stan' at line 2; included from 'model_continuous' at line 56)
#> failed to create the sampler; sampling not done
#> Error in check_stanfit(stanfit) : 
#>   Invalid stanfit object produced please report bug

etc…

But these, I think, are due to the fact that there is only one fixed effect, so can’t have a vector of df, location, scale (mismatch in dimension declared and found - expecting 1, found 2).

The original error is because the length of num_normals should be 1, but is 0. But I have no idea what num_normals is…

6

We’re surely misunderstanding something here; better bring in some heavy artillery :) @bgoodri @jonah?

1 Like