Check whether a fitted FIMS model is ready to use

Prints the result of each convergence check that fit_fims() stored on the fit, similar to sdmTMB::sanity(). The checks and thresholds are the ones fit_fims() warns with:

• optimizer: the nlminb convergence code is 0.

• max_gradient: the maximum absolute gradient is below 0.001. It fails above 0.01, and then sdreport is skipped.

• nll: the objective, the total negative log-likelihood, and each negative log-likelihood component are finite.

• hessian: the Hessian is positive definite.

• standard_errors: no standard error is NA.

• large_standard_errors: no fixed or random effect has a standard error above 1000.

• condition_number: the condition number of the Hessian is at or below 1e5.

The checks are read from the fit rather than recomputed, so this works after clear().

Usage

sanity_fims(fit)

Arguments

fit

An object of class FIMSFit returned by fit_fims().

Value

Invisibly, a list with all_ok, which is TRUE when every check is "OK", and checks, the tibble returned by get_convergence().

Examples

if (FALSE) { # \dontrun{
data("data_big")
data_4_model <- FIMSFrame(data_big)
fit <- setup_default_parameters(data = data_4_model) |>
  initialize_fims(data = data_4_model) |>
  fit_fims(optimize = TRUE)
sanity <- sanity_fims(fit)
sanity[["all_ok"]]
} # }