User defined formulas in bf()

Interfaces
1

I’m trying to use a rather simple two piece linear model (which makes it non-linear) to count data. However, I can’t figure out how to specify it correctly.

The function I wrote is

 two_piece <- function(x, x0, y0, xmax = 46, y_xmax = 0) {
                   delta_x <- xmax - max(x0, x) # val1 or val2
                   b0 = (y_xmax-y0)/(xmax - x0)
                   mu = y_xmax  - delta_x * b0
                   return(mu)
}

But when I try using

fit <- brm(
               bf(two_piece(x, x0, y0, xmax = 46, y_xmax = 0),
               shape ~ 1,
               x0 ~ 1, # + 1|xx,
               y0 ~ 1, # + 1|yy,
               nl = TRUE),
     data = df,
     family = negbinomial(link = "identity", link_shape = "identity"),
          prior =  prior(normal(150, 200), nlpar = "y0") +
                      prior(uniform(25, 45.5), 
                               lb = 25, ub = 45.5, nlpar = "x0"),
    save_model = "brms_two.piece.stan")

I get the error

Compiling Stan program...
Error in stanc(file = file, model_code = model_code, model_name = model_name,  : 
  0

Syntax error in 'string', line 42, column 61 to column 62, parsing error:

Ill-formed expression. Found identifier. There are many ways to complete this to a well-formed expression.

Where line 42 in brms_two.piece.stan has

mu[n] = two_piece(C_1[n] , nlp_x0[n] , nlp_y0[n] , xmax=46 , y_xmax=0);

So, I infer that stan cannot see the definition of two_piece, which makes some sense since it’s an R function. However, what doesn’t make sense to me is how one might, if at all, use a user defined function within brms()

(Note that I have tried defining the formula directly in bf(), but that leads to its own problems that I am posting about separately.)

  • Operating System: Ubuntu 22.04
  • Interface Version:
    • R version 4.2.2 Patched (2022-11-10 r83330)
    • brms_2.18.0
    • rstan_2.26.13
  • Compiler/Toolkit:
2

Man this is embarrassing. I asked a similar question here. Which essentially addresses this question.