Specifying different priors for different coefficients in RStanARM?

Interfaces rstanarm
1

Is there a way to specify different priors in RStanARM?
In Stan, you can do x1 ~ normal(0,1), x2 ~ normal(0,3), etc…
I do not see a way to do that in RStanARM- only prior, prior_intercept and prior_aux.

2

It is like

prior = normal(location = 0, scale = c(1, 3))

perhaps with auto_scale = FALSE.

3

Yeah it’s like @bgoodri said.

kickstarter_priors ← normal(location = c(-1, 0.5, 0.5),
scale = c(0.5, 0.5, 0.5), autoscale = FALSE)

1 Like
4

I presume, in general, parameters have to be sorted alphabetically in order to related them to coefficients? It is not possible to provide named vectors or lists for location or scale?

5

They variables don’t have to be sorted alphabetically, but the priors do take vectors only (possibly with recycling). So, the vectors different locations and / or scales have to be in the same order as the predictors in the model.

1 Like
6

Sorry, what you mean by „in the same order as the predictors in the model“? Would this mean that in a model

outcome ~ Z + Y + X + (W | ID)

I would specify (say) in the location vector the location for Z,Y,X at the 1st, 2nd and 3rd position, respectively?

7

Yes

2 Likes