quap allows incorrect(?) definition with categorical variable

[michaldanaj]

In mu definition, I omtited the index variable. Quap accepted it and gave results substantially different than with indexes. I think it's a bug, as there are no alpha and beta parameters defined, but alpha_S and beta_S.

  library(rethinking)
  data("Howell1")
  d <- Howell1[Howell1$age < 13,]
  d$S <- d$male + 1

  m2.3<- quap(
    alist(
      weight ~ dnorm(mu, sigma),
      mu <- alpha + beta * age,
      #mu <- alpha[S] + beta[S] * age,
      sigma ~ dexp(1/10),

      alpha[S] ~ dnorm(3,4),
      beta[S] ~ dunif(0,10)

    ), data = d
    ,start=list(alpha=10, beta=3, sigma=10)
  )

posterior2.3 <- precis(m2.3, depth = 2)
posterior2.3

[rmcelreath]

Seems to work as intended, in the sense that you defined vectors for alpha and beta and then used the vectors in the linear model. This means that "alpha" is the name if the whole vector, which you declared in the model code.