Skip to contents

Regression coefficient tables

Source: vignettes/regression.Rmd
regression.Rmd

Overview

tbl_regression() formats a tidy data frame of regression results — β, SE, 95% CI, and p-value — into a publication-ready table. It is designed to accept the output of broom::tidy() directly.

Column-name arguments take quoted strings. Defaults match broom::tidy(model, conf.int = TRUE) output:

Argument Default broom::tidy() column
term "term" model term
estimate "estimate" coefficient β
std_error "std.error" standard error
conf_low "conf.low" lower CI bound
conf_high "conf.high" upper CI bound
p "p.value" p-value

clerk_reg_example already uses these column names:

clerk_reg_example
#>          term estimate std.error conf.low conf.high p.value
#> 1         bmi    0.021     0.023   -0.024     0.066   0.012
#> 2       waist   -0.018     0.024   -0.065     0.029   0.048
#> 3 systolic_bp    0.012     0.011   -0.010     0.034   0.210
#> 4         bdi    0.008     0.022   -0.035     0.051   0.330
#> 5   panas_neg    0.015     0.018   -0.020     0.050   0.041
#> 6         age    0.003     0.015   -0.026     0.032   0.610
#> 7     sexMale   -0.045     0.020   -0.084    -0.006   0.003

Basic usage

Since clerk_reg_example matches all defaults, no remapping is needed:

tbl_regression(
  clerk_reg_example,
  output = "gt"
) |>
  clerk_render(title = "Linear regression: TMT completion time (log s)")
Linear regression: TMT completion time (log s)
β SE 95% CI p
All variables
bmi +0.021 0.023 [-0.024, +0.066] 0.012
waist -0.018 0.024 [-0.065, +0.029] 0.048
systolic_bp +0.012 0.011 [-0.010, +0.034] 0.210
bdi +0.008 0.022 [-0.035, +0.051] 0.330
panas_neg +0.015 0.018 [-0.020, +0.050] 0.041
age +0.003 0.015 [-0.026, +0.032] 0.610
sexMale -0.045 0.020 [-0.084, -0.006] 0.003

Domain grouping 

tbl_regression(
  clerk_reg_example,
  domains = list(
    "Cardiometabolic" = c("bmi", "waist", "systolic_bp"),
    "Mental health"   = c("bdi", "panas_neg")
  ),
  output = "gt"
) |>
  clerk_render(title = "Linear regression: TMT completion time (log s)")
#> Warning in `[<-.factor`(`*tmp*`, is.na(tbl[["domain"]]), value = "Other"):
#> invalid factor level, NA generated
Linear regression: TMT completion time (log s)
β SE 95% CI p
Cardiometabolic
bmi +0.021 0.023 [-0.024, +0.066] 0.012
waist -0.018 0.024 [-0.065, +0.029] 0.048
systolic_bp +0.012 0.011 [-0.010, +0.034] 0.210
Mental health
bdi +0.008 0.022 [-0.035, +0.051] 0.330
panas_neg +0.015 0.018 [-0.020, +0.050] 0.041
NA
age +0.003 0.015 [-0.026, +0.032] 0.610
sexMale -0.045 0.020 [-0.084, -0.006] 0.003

FDR correction 

tbl_regression(
  clerk_reg_example,
  domains = list(
    "Cardiometabolic" = c("bmi", "waist", "systolic_bp"),
    "Mental health"   = c("bdi", "panas_neg")
  ),
  fdr    = TRUE,
  output = "gt"
) |>
  clerk_render(
    title    = "Linear regression: TMT completion time (log s)",
    footnote = "FDR correction applied across all terms (BH)."
  )
#> Warning in `[<-.factor`(`*tmp*`, is.na(tbl[["domain"]]), value = "Other"):
#> invalid factor level, NA generated
Linear regression: TMT completion time (log s)
β SE 95% CI p p (FDR)
Cardiometabolic
bmi +0.021 0.023 [-0.024, +0.066] 0.012 0.042
waist -0.018 0.024 [-0.065, +0.029] 0.048 0.084
systolic_bp +0.012 0.011 [-0.010, +0.034] 0.210 0.294
Mental health
bdi +0.008 0.022 [-0.035, +0.051] 0.330 0.385
panas_neg +0.015 0.018 [-0.020, +0.050] 0.041 0.084
NA
age +0.003 0.015 [-0.026, +0.032] 0.610 0.610
sexMale -0.045 0.020 [-0.084, -0.006] 0.003 0.021
FDR correction applied across all terms (BH).

Using broom::tidy() directly 

library(broom)

lm(log(tmt_time) ~ bmi + waist + systolic_bp + bdi + panas_neg + age + sex,
   data = clerk_example) |>
  tidy(conf.int = TRUE) |>
  tbl_regression(
    domains = list(
      "Cardiometabolic" = c("bmi", "waist", "systolic_bp"),
      "Mental health"   = c("bdi", "panas_neg")
    ),
    fdr    = TRUE,
    output = "gt"
  ) |>
  clerk_render(title = "Linear regression: TMT completion time (log s)")

Logistic regression — exponentiated coefficients

Set exponentiate = TRUE to show odds ratios or hazard ratios:

library(broom)

glm(outcome ~ bmi + age + sex, data = my_data, family = binomial) |>
  tidy(conf.int = TRUE, exponentiate = FALSE) |>
  tbl_regression(
    exponentiate = TRUE,
    output       = "gt"
  ) |>
  clerk_render(
    title    = "Logistic regression: odds ratios",
    footnote = "Values shown as OR [95% CI]."
  )

Custom column names

Pass column names as strings when they differ from the defaults:

my_results <- data.frame(
  predictor = c("bmi", "age"),
  beta      = c(0.021, 0.003),
  se        = c(0.008, 0.002),
  lo95      = c(0.005, -0.001),
  hi95      = c(0.037,  0.007),
  pval      = c(0.012,  0.610)
)

tbl_regression(
  my_results,
  term      = "predictor",
  estimate  = "beta",
  std_error = "se",
  conf_low  = "lo95",
  conf_high = "hi95",
  p         = "pval",
  output    = "gt"
) |>
  clerk_render(title = "Regression results")

HTML output 

tbl_regression(
  clerk_reg_example,
  output = "html"
) |>
  clerk_render()