Tidy a beezdiscounting_tmb model into a coefficient tibble

Returns fixed-effect rows (the log-k coefficients) and/or variance-component rows, following the broom coefficient-table contract.

Usage

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

Arguments

x

A beezdiscounting_tmb object.

effects

Character vector: "fixed" (log-k fixed-effect rows), "ran_pars" (the RE SD and the auxiliary precision/scale parameter), or both (default).

report_space

Reporting scale for fixed-effect estimate/std.error: "natural" (default), "log10", "internal", or "log" ("internal" and "log" coincide for the log-k beta_k rows).

...

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"; the shape-parameter row (2-parameter equations only) carries 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. statistic and p.value are always computed on the estimation (log-k) scale. Wald statistics are not recomputed after back-transforming (broom convention; see the summary() note for details). Each test is of the log-scale coefficient against 0: for the intercept row that is k = 1 (rarely a meaningful null), and for a condition or covariate row it is a fold-change of 1 (no difference). The s row tests s = 1. Variance-component rows carry NA for statistic and p.value and are not affected by report_space.