Feature Comparison

In summata: Publication-Ready Summary Tables and Forest Plots

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  eval = TRUE
)

This article compares summata with other R packages for regression table generation and summary statistics.

---

Similar Packages

The R ecosystem includes several well-established packages for creating publication-ready regression tables. The following comparison identifies areas of overlap and distinction between summata and its alternatives.

| Package | Primary Focus | |:--------|:--------------| | summata | Fast regression workflows, forest plots, multivariate regression | | gtsummary | Comprehensive table generation, gt ecosystem, maximum flexibility | | finalfit | Clinical research, missing data handling, bootstrap simulations | | arsenal | Large-scale summaries, SAS-like output | | stargazer | Econometrics, LaTeX output, academic journal formatting | | tableone | Simple Table 1 generation, SMD calculations | | compareGroups | Bivariate analysis, clinical epidemiology |

---

Core Feature Matrix

# Define the feature matrix
features <- data.frame(
  Feature = c(
    "Descriptive Tables",
    "Stratified Summaries", 
    "Univariable Screening",
    "Multivariable Workflow",
    "Multivariate Regression",
    "Model Comparison",
    "Mixed-Effects Models",
    "Cox/Survival Models",
    "Interaction Formatting",
    "Forest Plots",
    "Table Merge/Stack",
    "Export (Word/PDF)",
    "Variable Labels"
  ),
  summata = c("Yes", "Yes", "Yes", "Yes", "Yes", "Yes", "Yes", "Yes", "Yes", "Yes", "No", "Yes", "Yes"),
  gtsummary = c("Yes", "Yes", "Yes", "Yes", "No", "Yes", "Yes", "Yes", "Partial", "Partial", "Yes", "Yes", "Yes"),
  finalfit = c("Yes", "Yes", "Yes", "Yes", "No", "Partial", "Partial", "Yes", "Partial", "Yes", "Partial", "Yes", "Yes"),
  arsenal = c("Yes", "Yes", "Yes", "Yes", "No", "No", "No", "Yes", "Partial", "No", "No", "Yes", "Yes"),
  stargazer = c("Yes", "No", "No", "Yes", "No", "Yes", "Partial", "Yes", "No", "No", "No", "Yes", "Partial"),
  tableone = c("Yes", "Yes", "No", "No", "No", "No", "No", "No", "No", "No", "No", "Partial", "Yes"),
  compareGroups = c("Yes", "Yes", "Yes", "No", "No", "No", "No", "Yes", "No", "No", "No", "Yes", "Yes"),
  stringsAsFactors = FALSE
)

# Create HTML table with color coding
cat('<table class="table table-sm" style="font-size: 0.9em;">\n')
cat('<thead>\n<tr>\n')
cat('<th style="text-align: left;">Feature</th>\n')
cat('<th style="text-align: center;">summata</th>\n')
cat('<th style="text-align: center;">gtsummary</th>\n')
cat('<th style="text-align: center;">finalfit</th>\n')
cat('<th style="text-align: center;">arsenal</th>\n')
cat('<th style="text-align: center;">stargazer</th>\n')
cat('<th style="text-align: center;">tableone</th>\n')
cat('<th style="text-align: center;">compareGroups</th>\n')
cat('</tr>\n</thead>\n<tbody>\n')

for (i in 1:nrow(features)) {
  cat('<tr>\n')
  cat(sprintf('<td style="text-align: left;">%s</td>\n', features$Feature[i]))

  for (pkg in c("summata", "gtsummary", "finalfit", "arsenal", "stargazer", "tableone", "compareGroups")) {
    val <- features[[pkg]][i]
    if (val == "Yes") {
      style <- 'background-color: #d4edda; color: #155724; text-align: center;'
      symbol <- "✓"
    } else if (val == "Partial") {
      style <- 'background-color: #fff3cd; color: #856404; text-align: center;'
      symbol <- "◐"
    } else {
      style <- 'background-color: #f8d7da; color: #721c24; text-align: center;'
      symbol <- "—"
    }
    cat(sprintf('<td style="%s">%s</td>\n', style, symbol))
  }
  cat('</tr>\n')
}

cat('</tbody>\n</table>\n')
cat('<p style="font-size: 0.85em; color: #666;"><strong>Legend:</strong> ✓ Full support | ◐ Partial support | — Not available</p>\n')

---

Feature Definitions

| Feature | Description | |:--------|:------------| | Descriptive Tables | Summary statistics tables (mean, SD, median, IQR, n, %) | | Stratified Summaries | Table 1-style summaries stratified by group with p-values | | Univariable Screening | Test multiple predictors against one outcome (crude associations) | | Multivariable Workflow | Combined univariable + multivariable analysis in one table | | Multivariate Regression | Test one predictor across multiple outcomes | | Model Comparison | Compare multiple models side-by-side with fit statistics | | Mixed-Effects Models | Support for lmer/glmer/coxme random effects models | | Cox/Survival Models | Cox proportional hazards and survival analysis | | Interaction Formatting | Native support for interaction terms with formatted output | | Forest Plots | Integrated forest plot visualization from regression results | | Table Merge/Stack | Combine separate tables horizontally or vertically | | Export (Word/PDF) | Direct export to publication formats | | Variable Labels | Apply custom labels to variables in output |

---

Unique Strengths of summata

The following features distinguish summata from comparable packages:

Multivariate Regression

The multifit() and multiforest() functions implement an inverted screening paradigm: testing a single predictor across multiple outcomes simultaneously. This workflow is common in epidemiological and clinical research but not directly supported by other packages.

# Test treatment effect across multiple outcomes
result <- multifit(
  data = clintrial,
  outcomes = c("surgery", "pfs_status", "os_status"),
  predictor = "treatment",
  covariates = c("age", "sex", "stage")
)

Integrated Forest Plots

Customizable forest plots are generated directly from analysis results without intermediate steps:

# From univariable screening
screen_result <- uniscreen(data, outcome, predictors)
uniforest(screen_result)

# From multivariate regression
multi_result <- multifit(data, outcomes, predictor)
multiforest(multi_result)

Performance Optimization

Built on data.table for computational efficiency, summata provides competitive performance across all benchmarked workflows. With conf_method = "wald", it is the fastest option tested for GLM-based regression tables and univariable screening. With the default profile likelihood CIs, performance is comparable to other packages that use the same CI method (finalfit, broom). See the Benchmarks article for detailed comparisons.

Unified API

All modeling functions share consistent syntax across model types:

# Same interface for different model types
fit(data, outcome, predictors, model_type = "glm")
fit(data, outcome, predictors, model_type = "coxph")
fit(data, outcome, predictors, model_type = "lmer", random = "(1|site)")

Complete Mixed-Effects Support

Full support for coxme (mixed-effects Cox models) alongside lmer and glmer, which is limited or absent in other packages.

---

Additional Resources

• Benchmarks — Performance comparisons
• gtsummary documentation
• finalfit documentation
• arsenal documentation

summata documentation built on May 7, 2026, 5:07 p.m.