Build newdata for the meta model

The meta model holds individual level rows and summary rows in one data frame, telling them apart by the obs_type slot. Prediction is for the delay distribution itself, so this method builds an individual level row. It takes the same arguments as epidist_newdata.epidist_marginal_model(), and fills the slots a summary row would use with the values an individual row carries. The result is what delay_parameter_draws() and add_summaries() need, so there is no need to copy a summary row out of the model data and overwrite its covariates.

Usage

# S3 method for class 'epidist_meta_model'
epidist_newdata(
  data,
  ...,
  pwindow = 0,
  swindow = 0,
  relative_obs_time = Inf,
  delay_min = 0
)

Arguments

data

An epidist data object, such as one returned by as_epidist_marginal_model(), as_epidist_latent_model(), as_epidist_naive_model() or as_epidist_meta_model().

...

Variables to expand into a grid, passed to tidyr::expand(). Supply the variables used in the model formula, such as sex. Each combination of their unique values becomes a row. Supply no variables to get a single row, which is what an intercept only model needs. A variable expanded here keeps its expanded values, so naming it as an argument of the method as well is an error.

pwindow

Width of the primary event censoring window. Defaults to 0, which is no censoring.

swindow

Width of the secondary event censoring window. Defaults to 0, which is no censoring.

relative_obs_time

Observation time relative to the primary event. Defaults to Inf, which is no right truncation. brms warns about infinite values in the data when this is Inf. That warning is safe here, because prediction is done in R and never passes the value to Stan.

delay_min

Minimum delay, the left truncation point. Defaults to 0, which is no left truncation.

Value

A tibble::tibble() of newdata ready to predict from.

Details

A model with a study level term, such as mu ~ 1 + (1 | study), needs either a study column in newdata or re_formula = NA when predicting. Expand study here to predict for each study, or leave it out and pass re_formula = NA for the population level delay. The primary event distribution of the individual level rows is a parameter of the model rather than a column, so it needs nothing here.

See also

Other meta_model: as_epidist_meta_model(), as_epidist_meta_model.NULL(), as_epidist_meta_model.epidist_aggregate_data(), as_epidist_meta_model.epidist_estimates_data(), as_epidist_meta_model.epidist_linelist_data(), assert_epidist.epidist_meta_model(), epidist_family_model.epidist_meta_model(), epidist_formula_model.epidist_meta_model(), epidist_model_prior.epidist_meta_model(), epidist_transform_data_model.epidist_meta_model(), is_epidist_meta_model(), new_epidist_meta_model()

Other newdata: epidist_newdata(), epidist_newdata.default(), epidist_newdata.epidist_latent_model(), epidist_newdata.epidist_marginal_model(), epidist_newdata.epidist_naive_model()

Examples

estimates <- as_epidist_estimates_data(
  data.frame(
    study = c("A", "A", "B"),
    type = c("mean", "sd", "mean"),
    value = c(7.5, 3.6, 6.4),
    n = c(120, 120, 80),
    relative_obs_time = c(20, 20, Inf),
    trunc_adjusted = c(FALSE, FALSE, TRUE),
    cens_adjusted = c(0, 0, 1)
  )
)
#> ℹ No `pwindow` column supplied, assuming a censoring window of 1 (daily
#>   reporting) for every study.
#> ℹ No `swindow` column supplied, assuming a censoring window of 1 (daily
#>   reporting) for every study.
#> ℹ No trunc_design column supplied, assuming every study that did not adjust for
#>   right truncation followed a cohort with a common observation time rather than
#>   accruing primary events up to a calendar collection stop.
#> ℹ No max_delay column supplied, using the delay beyond which 1% of the second
#>   moment of a lognormal matched to each study's summaries lies (at least 10 and
#>   at most twenty times the largest reported value, in whole secondary windows)
#>   as the grid cutoff, or five times the largest reported value where nothing
#>   can be matched. Raise it if the delay has a longer tail than that, and lower
#>   it to speed up fitting.
meta <- as_epidist_meta_model(estimates = estimates)

# The population level delay, with no censoring and no truncation
epidist_newdata(meta)
#> # A tibble: 1 × 18
#>   delay_lwr obs_type study_n trunc_adjusted trunc_design cens_adjusted
#>       <dbl>    <int>   <int>          <int>        <int>         <int>
#> 1         0        1       0              0            0             0
#> # ℹ 12 more variables: group_start <int>, group_len <int>, chol_start <int>,
#> #   n_quad <int>, relative_obs_time <dbl>, pwindow <dbl>, swindow <dbl>,
#> #   delay_upr <dbl>, delay_min <dbl>, report_se <dbl>, quantile_p <dbl>,
#> #   growth_rate <dbl>

# A row for each study, with daily censoring
epidist_newdata(meta, study, pwindow = 1, swindow = 1)
#> # A tibble: 2 × 19
#>   study delay_lwr obs_type study_n trunc_adjusted trunc_design cens_adjusted
#>   <chr>     <dbl>    <int>   <int>          <int>        <int>         <int>
#> 1 A             0        1       0              0            0             0
#> 2 B             0        1       0              0            0             0
#> # ℹ 12 more variables: group_start <int>, group_len <int>, chol_start <int>,
#> #   n_quad <int>, relative_obs_time <dbl>, pwindow <dbl>, swindow <dbl>,
#> #   delay_upr <dbl>, delay_min <dbl>, report_se <dbl>, quantile_p <dbl>,
#> #   growth_rate <dbl>