Tidy a beezdemand_tmb Model

Usage

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

Arguments

x: A beezdemand_tmb object.
effects: Character. Which effects to return: "fixed" for the fixed-effect (core demand parameter) rows, "ran_pars" for the random-effect variance components, or both (the default). Matches the effects argument of tidy.beezdemand_nlme().
report_space: Character. Reporting space for the fixed-effect (core demand parameter) rows. One of "natural", "log10", or "internal". Variance-component rows are unaffected (see Details). estimate/std.error follow this scale; statistic/p.value are always on the estimation scale (transformation-invariant).
...: Additional arguments.

Value

A tibble of model terms with columns term, estimate, std.error, statistic, p.value, component, estimate_scale, and term_display. An estimate_internal column (the pre-transform estimate) is additionally present whenever effects includes "fixed". Fixed-effect rows carry component == "fixed" (matching tidy.beezdemand_nlme() and the nlme/lme4 convention); variance-component rows carry component == "variance". A hessian_warning attribute (character scalar, or absent) is attached depending on x$hessian_pd: absent (no attribute) when hessian_pd is TRUE or NULL (the field is missing on a legacy fit object – nothing to say); a message noting the Hessian is not positive definite when hessian_pd is FALSE; a message noting positive-definiteness is unknown (because TMB::sdreport() failed entirely, so SEs/CIs are unavailable, not merely unreliable) when hessian_pd is NA. This attribute is not printed by an ordinary tibble print – see summary.beezdemand_tmb() or check_demand_model() for the surfaced versions of the same diagnostic.

Details

Variance-component rows (effects = "ran_pars") are exactly the rows of summary(x)$variance_components: the Q0 and alpha random-effect standard deviations on the log10 scale and the residual standard deviation on the model's likelihood scale. They are not the raw internal logsigma optimizer coefficients and do not respond to report_space; std.error is NA for them. Random-effect correlations are not tidied here – see summary(x)$correlations or VarCorr(x) for those. The NLME sibling tidy.beezdemand_nlme() likewise reports SDs, so backend-agnostic code can consume the estimate column without dispatch logic on either side.

Examples

# \donttest{
data(apt)
fit <- fit_demand_tmb(apt, equation = "exponential", verbose = 0)
#>   equation='exponential': Dropped 14 zero-consumption observations (146 remaining).
tidy(fit)
#> # A tibble: 6 × 9
#>   term           estimate std.error statistic   p.value component estimate_scale
#>   <chr>             <dbl>     <dbl>     <dbl>     <dbl> <chr>     <chr>         
#> 1 Q0:(Intercept)  6.51      0.810       15.1   2.59e-51 fixed     natural       
#> 2 alpha:(Interc…  0.00302   0.00169    -10.4   3.75e-25 fixed     natural       
#> 3 log_k           0.895     0.484        1.85  6.42e- 2 fixed     log           
#> 4 sigma_b (Q0 R…  0.167    NA           NA    NA        variance  log10         
#> 5 sigma_c (alph…  0.199    NA           NA    NA        variance  log10         
#> 6 sigma_e (Resi…  0.142    NA           NA    NA        variance  natural       
#> # ℹ 2 more variables: term_display <chr>, estimate_internal <dbl>
tidy(fit, effects = "fixed", report_space = "log10")
#> # A tibble: 3 × 9
#>   term            estimate std.error statistic  p.value component estimate_scale
#>   <chr>              <dbl>     <dbl>     <dbl>    <dbl> <chr>     <chr>         
#> 1 Q0:(Intercept)     0.814    0.0540     15.1  2.59e-51 fixed     log10         
#> 2 alpha:(Interce…   -2.52     0.243     -10.4  3.75e-25 fixed     log10         
#> 3 log_k              0.895    0.484       1.85 6.42e- 2 fixed     log           
#> # ℹ 2 more variables: term_display <chr>, estimate_internal <dbl>
tidy(fit, effects = "ran_pars")
#> # A tibble: 3 × 8
#>   term             estimate std.error statistic p.value component estimate_scale
#>   <chr>               <dbl>     <dbl>     <dbl>   <dbl> <chr>     <chr>         
#> 1 sigma_b (Q0 RE …    0.167        NA        NA      NA variance  log10         
#> 2 sigma_c (alpha …    0.199        NA        NA      NA variance  log10         
#> 3 sigma_e (Residu…    0.142        NA        NA      NA variance  natural       
#> # ℹ 1 more variable: term_display <chr>
# }