Skip to contents

statsguider

pkgdown R-CMD-check License: MIT

statsguider is an R package for guided statistical test selection from a data.frame. It focuses on five jobs:

  • classify a simple group-comparison design
  • validate whether a simple test is appropriate
  • recommend a test from a small set of design choices
  • run supported base-R tests
  • redirect to model-based workflows when a simple test is the wrong branch

https://dai540.github.io/statsguider/

The input contract is intentionally simple:

  • data: the analysis table
  • outcome: the outcome column
  • group: the group or condition column when needed
  • id: the subject identifier for paired or repeated designs
  • choice-style arguments: goal, outcome_type, paired, repeated, adjust, normality, run, and language

Version 1.0.0 is intentionally narrow. The package is centered on simple single-outcome group comparison and explicitly redirects users when the data belong in regression, survival analysis, agreement analysis, equivalence, or count-model workflows.

Installation

Install from GitHub with pak:

install.packages("pak")
pak::pak("dai540/statsguider")

or remotes:

install.packages("remotes")
remotes::install_github("dai540/statsguider")

Then load the package:

Minimal Example

select_test() is the main wrapper:

dat <- subset(statsguider::wet_example, visit == "week4")

result <- statsguider::select_test(
  data = dat,
  outcome = "biomarker",
  group = "group",
  outcome_type = "continuous",
  paired = "no",
  repeated = "no",
  run = "recommend",
  language = "en"
)

result

To run the supported test immediately:

statsguider::select_test(
  data = dat,
  outcome = "biomarker",
  group = "group",
  outcome_type = "continuous",
  paired = "no",
  repeated = "no",
  run = "run",
  language = "en"
)

Main Functions

select_test() and guided_test() return either a statsguider_decision object or a statsguider_result object, depending on whether the user asks for recommendation only or full execution.

Supported Simple Testing Branches

Continuous outcomes

  • independent 2 groups: Welch t-test or Mann-Whitney U test
  • paired 2 groups: paired t-test or Wilcoxon signed-rank test
  • independent 3+ groups: Welch ANOVA or Kruskal-Wallis test
  • repeated 3+ groups: repeated-measures ANOVA or Friedman test

Categorical and ordinal outcomes

  • independent categorical comparison: chi-squared test or Fisher exact test
  • paired binary comparison: McNemar test
  • ordinal comparison: Mann-Whitney U, Wilcoxon signed-rank, Kruskal-Wallis, or Friedman test

Redirected branches

  • adjusted analyses
  • association questions
  • time-to-event outcomes
  • agreement and reproducibility
  • equivalence and non-inferiority
  • count outcomes

Built-in Example Data

  • wet_example

wet_example is a small wet-lab style dataset with continuous, binary, and ordinal outcomes across treatment groups and visits.

Tutorials

The website is organized around:

  • core introductions: start page, main functions, branching guide
  • English scenario tutorials
  • Japanese tutorials

Documentation

Citation

If you use statsguider, cite the package as:

Dai (2026). statsguider: Guided Statistical Test Selection from a Data Frame. R package. https://dai540.github.io/statsguider/

You can also retrieve the citation from R:

citation("statsguider")