Beginner-Friendly Statistics and Publication-Ready Tables

Provides beginner-friendly functions for common descriptive and inferential statistical analyses, together with tools for creating publication-ready tables. The package includes data description, summary statistics, distribution assessment, group comparisons, effect sizes, correlations, proportions and rates with confidence intervals, crosstabs for epidemiological measures, plots, and modular table-building workflows. Statistical methods include unequal-variance comparisons described by Welch (1947) , score confidence intervals described by Wilson (1927) , and robust variance assessment described by Brown and Forsythe (1974) .

Publication-ready descriptive and inferential statistics

gtstats helps students, researchers, and public-health analysts move from a new dataset to clear descriptive results, focused comparisons, and a publication-ready Table 1. It keeps the R syntax approachable while retaining the underlying results, denominators, assumptions, and automatic decisions for review.

Why it exists

Statistics packages often make a user choose between an oversimplified default and a highly technical workflow. gtstats is designed to bridge that gap:

• publication-ready tables are the default, not an extra formatting task;
• automatic test selection is visible and can always be overridden;
• Welch methods are the conservative default for suitable independent continuous comparisons; var_equal = TRUE is available when equal variances are prespecified and justified, not inferred from a variance test;
• distribution, variance, assumptions, and denominators are inspectable;
• the table builder offers simple defaults with optional layers of control.

The automatic algorithm is documented in full: marked skewness—not a Shapiro-Wilk p-value alone—triggers rank-based continuous tests; sparse categorical tables use Fisher’s exact test when an expected count is below 1 or more than 20% are below 5; and paired/repeated designs use their dedicated methods. See the inferential tests and assumptions guide.

gtstats uses established R statistical engines such as stats, dplyr, ggplot2, gt, and flextable; it provides a coherent teaching and reporting workflow around them.

Install

# Development version until gtstats is available on CRAN
remotes::install_github("ThinkDenominator/gtstats")
library(gtstats)

Built-in teaching data

gtstats includes five labelled datasets, so examples work without installing other packages:

data(
  "birthwt", "trial_data", "paired_data", "outbreak_data",
  "surveillance_data", package = "gtstats"
)

• birthwt: real low-birth-weight clinical data for descriptive tables, independent comparisons, and epidemiology.
• trial_data: a three-arm synthetic trial for Welch ANOVA, Kruskal-Wallis, chi-square, Fisher exact, correlations, and rates.
• paired_data: long-format paired follow-up data for paired t-tests, Wilcoxon signed-rank, and McNemar tests.
• outbreak_data: the real CDC Oswego foodborne-outbreak line list for attack rates and 2x2 epidemiology.
• surveillance_data: an archived CDC weekly hospital-admission extract for aggregate counts, population denominators, and rates per 100,000.

Prefer a guided interface?

The package remains code-first, but a Shiny companion is available when you prefer to learn by clicking through the workflow. It includes all five teaching datasets, data from the current R environment, CSV/Excel upload, a working data dictionary, distribution and variance checks, a layer-by-layer summary table, outbreak and surveillance tables, focused group comparisons, correlations and crosstabs. Tables can be downloaded as Word, HTML, PDF, or RTF, and every analysis has copyable and downloadable R code. Excel import uses the optional rio package.

install.packages("shiny") # once, if needed
install.packages("rio")   # once, only for Excel files
gtstats_app() # opens in the RStudio Viewer when available

The interface is intentionally a guide rather than a black box: use the code shown beside each result in an R script or Quarto document to keep the final analysis reproducible.

The app keeps three table routes separate. Choose Summary table for raw participant-level data, Epi table when events and denominators still need calculation, and Current data are final calculated results in Customise table only when every supplied row and value should be preserved exactly through as_stats_table().

Five-minute workflow

data("birthwt", package = "gtstats")

overview <- describe_data(birthwt)

distribution <- assess_distribution(birthwt, vars = c(age, lwt), by = low)
variance <- assess_variance(birthwt, vars = c(age, lwt), by = low)
# Use `test = "none"` for descriptive spreads only, or `test = "bartlett"`
# when the stronger normal-distribution assumption is justified.

table_one <- summary_table(
  birthwt,
  by = low,
  include = c(age, lwt, race, smoke),
  overall = "last",
  layout = "separate"
) |>
  add_ci(vars = c(age, race)) |>
  add_p()

# Optional categorical-only presentation with distinct n and % columns
categorical_table <- summary_table(
  birthwt,
  by = low,
  include = c(race, smoke),
  categorical_layout = "separate"
)

# Optional compact binary presentation: one clinically relevant event per row
compact_binary_table <- summary_table(
  birthwt,
  by = low,
  include = c(smoke, ht, race),
  show_dichotomous = "single_row",
  value = c(smoke = "Yes", ht = "Yes")
)

# Missing values are excluded from categorical percentage denominators by
# default. Use this explicit option only when Missing is itself a reported
# category and should contribute to 100%.
missing_as_category <- summary_table(
  birthwt,
  include = c(race, smoke),
  missing = "as_category"
)

comparison <- compare_groups(birthwt, variable = lwt, group = low)
comparison$method$selection_rule
diagnostics_stats(comparison)

Each object prints as a publication-ready flextable, ready for Word and PowerPoint. Use customise_table() to finish its appearance. Use to_gt() only when an HTML-focused gt table is specifically required.

table_one |>
  customise_table(
    theme = "journal",
    title = "Table 1. Maternal characteristics",
    spanning_header = list(
      "Birth-weight groups" = c(
        "low = Normal birth weight",
        "low = Low birth weight"
      )
    ),
    borders = "horizontal",
    pvalue_style = "threshold"
  ) |>
  save_output("table-1.docx", path = tempdir())

# Explicit HTML/gt route
to_gt(table_one, title = "Table 1. Maternal characteristics")

Browse by task

Publication tables and console output

Core statistical functions produce a publication-ready table by default. In a terminal, automated test, or plain-text workflow, request a real tibble:

compare_groups(birthwt_data, variable = age, group = low)

compare_groups(
  birthwt_data,
  variable = age,
  group = low,
  format = "tibble"
)

The same format = "table" or format = "tibble" choice is available for data description, distribution and variance assessment, effect sizes, correlations, proportions, rates, and cross-tabulations. A summary-table builder stays composable in either format, so add_p() and other layers still work before the completed tibble is printed.

Function map

correlation() handles both a prespecified pair and an exploratory matrix. The matrix retains pair-specific denominators and p-values in $summary, while plot_correlation() provides a labelled coefficient heatmap.

correlation(mtcars, vars = c(mpg, disp, hp, wt)) |>
  plot_correlation()

Citation

If gtstats contributes to your analysis, tables, teaching, or publication, please cite:

Polani R, Kaviprawin M, Sakthivel M, Eliyas SK, Krishnamoorthy Y (2026). gtstats: Beginner-Friendly Statistics and Publication-Ready Tables. R package version 1.0.0. https://gtstats.thinkdenominator.com/.

R returns the authoritative citation and BibTeX entry installed with the package:

citation("gtstats")
toBibtex(citation("gtstats"))

The pkgdown Citing gtstats article provides a copy-ready BibTeX record and reproducibility guidance. The reserved Zenodo DOI will be added here after its record is published and the DOI resolves publicly.