Compare univariate and multivariate models

This function builds both univariate models (each predictor separately) and a multivariate model (all predictors together), then combines the results for comparison. This is useful for understanding how predictor effects change when accounting for other variables.

Usage

br_compare_models(
  data,
  y,
  x,
  x2 = NULL,
  method,
  ...,
  n_workers = 1L,
  model_args = list(),
  run_args = list()
)

Arguments

data

A data.frame containing all necessary variables for analysis.

y

Character vector specifying dependent variables (response variables). For GLM models, this is typically a single character (e.g., "outcome"). For Cox-PH models, it should be a length-2 vector in the format c("time", "status").

x

Character vector specifying focal independent terms (predictors). These will be modeled both individually (univariate) and together (multivariate).

x2

Character vector specifying control independent terms (predictors, optional). These are included in all models (both univariate and multivariate).

method

Method for model construction. See br_set_model() for details.

...

Additional arguments passed to br_run().

n_workers

Integer, indicating number of workers for parallel processing.

model_args

A list of arguments passed to br_set_model().

run_args

A list of arguments passed to br_run().

Value

A list with class breg_comparison containing:

• univariate: breg object with univariate model results

• multivariate: breg object with multivariate model results

• combined_results: Combined results data frame with a mode column

• combined_results_tidy: Combined tidy results with a mode column

See also

Other br_compare: br_show_forest_comparison()

Examples

# Compare univariate vs multivariate for Cox models
lung <- survival::lung |>
  dplyr::filter(ph.ecog != 3)
lung$ph.ecog <- factor(lung$ph.ecog)

comparison <- br_compare_models(
  lung,
  y = c("time", "status"),
  x = c("ph.ecog", "ph.karno", "pat.karno"),
  x2 = c("age", "sex"),
  method = "coxph"
)
#> Building univariate models...
#> exponentiate estimates of model(s) constructed from coxph method at default
#> Building multivariate model...
#> exponentiate estimates of model(s) constructed from coxph method at default

# View combined results
comparison$combined_results_tidy
#> # A tibble: 4 × 9
#>   Focal_variable term    estimate std.error statistic p.value conf.low conf.high
#>   <chr>          <chr>      <dbl>     <dbl>     <dbl>   <dbl>    <dbl>     <dbl>
#> 1 ph.karno       ph.kar…    0.988   0.00595     -2.06 3.96e-2    0.976     0.999
#> 2 pat.karno      pat.ka…    0.981   0.00568     -3.30 9.64e-4    0.971     0.992
#> 3 ph.karno       ph.kar…    1.02    0.00984      1.59 1.12e-1    0.996     1.04 
#> 4 pat.karno      pat.ka…    0.989   0.00723     -1.55 1.21e-1    0.975     1.00 
#> # ℹ 1 more variable: mode <chr>

# Create forest plot comparison
br_show_forest_comparison(comparison)