a few issues with distributional parameters

[jfree-man]

I have found a couple problems that need to be addressed for distributional parameters.

• Fitting character distributional parameters - See the second calibration example in example("fit_distr_params") where user specified distributional parameter sd_var received the default transformation but the variable name prevents values from being back transformed in the coefficient table.
  • This also affects setting priors on distributional parameters, see this vignette example here where my_var gets different transformations for each prior term in the objective function.
• Missing prior location parameter - an absent location parameter for a prior on a parameter doesn't generate a warning or error. See example below. The objective function is missing this prior term but parameters beta and distr_params_sd_beta still show up in the coefficient table.

spec = mp_tmb_library("starter_models", "sir", package = "macpan2")
data = mp_simulator(spec, 50, "infection") |> mp_trajectory()
cal = mp_tmb_calibrator(
    spec
  , data
  , traj = "infection"
  , par = list(beta = mp_normal(sd = mp_fit(0.1)))
  , default = list(beta = 0.25)
)
cal$simulator$tmb_model$obj_fn$obj_fn_expr

mp_optimize(cal)
mp_tmb_coef(cal)

[stevencarlislewalker]

I have a fix for "Missing prior location parameter". Testing