I’m using `brm()`

with `cmdstanr`

as its backend (i.e. I set `backend = "cmdstanr"`

in formulae) and trying to calculate the Bays Factor using `bayes_factor()`

comparing multiple models.

However, `bayes_factor()`

returns `Error: Backend 'rstan' is required for this method`

and no computation shows progress. Actually, I really need to use `cmdstanr`

to analyse more complicated data repetitively with speed, and I dis-prefer to using `rstan`

this time. Is there any way to compute BF with brms with cmdstanr backend?

MWE

```
## For sleepstudy data
library(lme4)
## Null model
fm0 <- brm(
Reaction ~ 0 + Intercept + (0 + Intercept + Days|Subject),
data = sleepstudy,
family = "normal",
prior =
c(
prior(normal(0, 1), class = b, coef = Intercept),
prior(normal(0, 1), class = sd),
prior(lkj(2), class = cor)
),
save_pars = save_pars(all = TRUE),
backend = "cmdstanr"
)
## Alternative model
fm1 <- brm(
Reaction ~ 0 + Intercept + Days + (0 + Intercept + Days|Subject),
data = sleepstudy,
family = "normal",
prior =
c(
prior(normal(0, 1), class = b, coef = Intercept),
prior(normal(0, 1), class = b, coef = Days),
prior(normal(0, 1), class = sd),
prior(lkj(2), class = cor)
),
save_pars = save_pars(all = TRUE),
backend = "cmdstanr"
)
## The following commands return an error suggesting us to use rstan
bayes_factor(bridge_sampler(fm1), bridge_sampler(fm0))
bayes_factor(fm1, fm0)
```

I’m running on:

- Operating System: Windows 10 x64 (build 18362)
- CmdStan Version: 0.1.3
- brms Version: 2.14.0

Any suggestion is appreciated.