Tidy a beezdiscounting_choice model into a coefficient tibble

Returns fixed-effect rows (the log-k coefficients), the shape rows (the choice-sensitivity gamma and, when present, the choice-bias beta0), and the variance-component row, following the broom coefficient-table contract.

Usage

# S3 method for class 'beezdiscounting_choice'
tidy(
  x,
  effects = c("fixed", "ran_pars"),
  report_space = c("natural", "log10", "internal", "log"),
  ...
)

Arguments

x

A beezdiscounting_choice object.

effects

Character vector: "fixed" (log-k fixed-effect rows + the shape rows gamma/beta0), "ran_pars" (the RE SD), or both (default).

report_space

"natural", "log10", "internal", or "log" - reporting scale for the fixed-effect estimate/std.error. Default is "natural".

...

Unused.

Value

A tibble with exactly 8 columns in this order: term, estimate, std.error, statistic, p.value, component, estimate_scale, term_display. Fixed-effect rows carry component == "fixed"; gamma/beta0 carry component == "shape"; variance rows carry component == "variance".

Details

estimate and std.error are reported on the report_space scale for the fixed-effect (beta_k) rows and for gamma (which is transformed EXPLICITLY since the param-space transformer keys only on k/s/phi). beta0 is on the identity (logit-intercept) scale and is NEVER transformed across report spaces. statistic and p.value are always computed on the estimation (internal) scale - Wald statistics are not recomputed after back-transforming (broom convention).