model.matrix() fails for stan_betareg if x = TRUE

Summary:

By default, x = FALSE in stan_betareg() in which case model.matrix() works. However, if x = TRUE, then model.matrix() eventually fails.

Description:

model.matrix() calls model.matrix.stanreg(), which hands off to model.matrix.betareg(), which is

function (object, model = c("mean", "precision"), ...) 
{
    model <- match.arg(model)
    rval <- if (!is.null(object$x[[model]])) 
        object$x[[model]]
    else model.matrix(object$terms[[model]], model.frame(object), 
        contrasts = object$contrasts[[model]])
    return(rval)
}

this falls to the else clause by default because object$x partially matches object$xlevels, which is a list without an element called "mean" (or "precision"). But when x = TRUE, in the call to stan_betareg, object$x matches this matrix, which cannot be indexed by [model].

In addition, if model = FALSE in the call to stan_betareg(), then model.matrix() will fail even if x = FALSE because when it falls into else model.matrix(...), model.matrix.default() tries to extract object$model, which partially matches object$modeling_function and barfs on this character string.

Reproducible Steps:

example(stan_betareg, ask = FALSE)
model.matrix(update(fit, x = TRUE))

RStanARM Version:

2.15.3

R Version:

3.4.0

Operating System:

Debian