`bayesian_as_frequentist()` fails for brms `0 + Intercept` formula

[DominiqueMakowski]

The 0 + Intercept syntax in brms is useful to facilitate prior specification for the intercept, but it leads to issues in check_model whereby it wrongly believes that Intercept is a variable and not a special term.

model <- brms::brm(qsec ~ mpg, data=mtcars, refresh=0)
#> Compiling Stan program...
#> Start sampling
performance::check_model(model)
#> Not enough model terms in the conditional part of the model to check for
#>   multicollinearity.

model <- brms::brm(qsec ~ 0 + Intercept + mpg, data=mtcars, refresh=0)
#> Compiling Stan program...
#> Start sampling
performance::check_model(model)
#> Error in eval(predvars, data, env): object 'Intercept' not found

^{Created on 2024-03-29 with reprex v2.0.2}

[strengejacke]

Error comes from bayestestR::bayesian_as_frequentist(model).

[DominiqueMakowski]

Should be fixed by making our own plot (see https://github.com/easystats/performance/issues/477)

[strengejacke]

Not sure. Most of current checks don't work for Bayesian models, that's why we convert to frequentist

[strengejacke]

@DominiqueMakowski bayesian_to_frequentist() is much better now, also capturing less common families. We should include this patch before CRAN submission (if not too late).