Logistic regression adjusted differences

This function works with gtsummary::add_difference() to calculate adjusted differences and confidence intervals based on results from a logistic regression model. Adjustment covariates are set to the mean to estimate the adjusted difference. The function uses bootstrap methods to estimate the adjusted difference between two groups. The CI is estimate by either using the SD from the bootstrap difference estimates and calculating the CI assuming normality or using the centiles of the bootstrapped differences as the confidence limits

The function can also be used in add_p(), and if you do, be sure to set boot_n = 1 to avoid long, unused computation.

logistic_reg_adj_diff(
  data,
  variable,
  by,
  adj.vars,
  conf.level,
  type,
  ci_type = c("sd", "centile"),
  boot_n = 250,
  ...
)

Arguments

data

a data frame

variable

string of binary variable in data=

by

string of the by= variable name

adj.vars

character vector of variable names to adjust model for

conf.level

Must be strictly greater than 0 and less than 1. Defaults to 0.95, which corresponds to a 95 percent confidence interval.

type

string indicating the summary type

ci_type

string dictation bootstrap method for CI estimation. Must be one of c("sd", "centile").

boot_n

number of bootstrap iterations to use. In most cases, it is reasonable to used 250 for the "sd" method and 5000 for the "centile" method.

...

not used

Value

tibble with difference estimate

Example Output

Example 1

Example 2

See also

Other gtsummary-related functions: add_inline_forest_plot(), add_sparkline(), as_ggplot(), bold_italicize_group_labels(), style_tbl_compact(), tbl_likert(), theme_gtsummary_msk()

Examples

library(gtsummary)
tbl <- tbl_summary(trial, by = trt, include  = response, missing = "no")

# Example 1 -----------------------------------------------------------------
logistic_reg_adj_diff_ex1 <-
  tbl %>%
  add_difference(
    test = everything() ~ logistic_reg_adj_diff,
    adj.vars = "stage"
  )

# Example 2 -----------------------------------------------------------------
# Use the centile method, and
# change the number of bootstrap resamples to perform
logistic_reg_adj_diff_ex2 <-
  tbl %>%
  add_difference(
    test = everything() ~
      purrr::partial(logistic_reg_adj_diff, ci_type = "centile", boot_n = 100),
    adj.vars = "stage"
  )