Skip to contents

statease

CRAN TotalCRAN Version

Statistical analysis with plain-English interpretation for R

Overview

statease runs common statistical tests and returns each result together with a plain-English interpretation, the effect size, the significance decision, and, as of v1.4.0, a set of assumption checks relevant to that specific test, shown by default rather than as an optional extra step.

statease does not replace statistical judgment. It cannot know your study design, whether your sample was randomly selected, or whether an assumption violation matters for your particular use case, no automated tool can. What it does is surface the diagnostic information a careful analyst would normally have to compute separately (normality, variance homogeneity, multicollinearity, and several others depending on the test), clearly labelled as PASSED, WARNING, or NOTE, so that information is in front of you at the moment you read the result rather than something you have to remember to go check yourself.

You can also describe your study design in a sentence, and statease will echo it back alongside the interpretation as a reminder to read the result in that context:

ttest_interpret(x, y, context = "observational sample, not randomized")

Installation 

install.packages("statease")

For the development version from GitHub:

# install.packages("devtools")
devtools::install_github("DevWebWacky/statease")

Live App

Try statease directly in your browser without installing R:

🌐 Launch statease Shiny App

Functions

Function What it does
analyze() Master function - auto-detects and runs the right test
describe() Descriptive statistics with interpretation
ttest_interpret() T-tests, with normality and variance checks by default
anova_interpret() One-way ANOVA with Tukey post-hoc, eta squared, and assumption checks
anova2_interpret() Two-way ANOVA with Type II/III SS and assumption checks
manova_interpret() MANOVA with Pillai’s trace and follow-up ANOVAs
chisq_interpret() Chi-square test with Cramer’s V and expected-frequency checks
fisher_interpret() Fisher’s Exact Test with Odds Ratio
mcnemar_interpret() McNemar’s Test for paired categorical data
cor_interpret() Correlation (Pearson, Spearman, Kendall) with linearity notes
reg_interpret() Simple linear regression with normality, homoscedasticity, and independence checks
mlr_interpret() Multiple linear regression, adding multicollinearity (VIF) checks
logistic_interpret() Logistic regression with odds ratios and a separation diagnostic
mannwhitney_interpret() Mann-Whitney U test (non-parametric)
wilcoxon_interpret() Wilcoxon Signed Rank test (non-parametric)
kruskal_interpret() Kruskal-Wallis test with post-hoc comparisons
friedman_interpret() Friedman Test with Kendall’s W
check_assumptions() Run the same assumption checks on their own, before choosing a test
power_interpret() Statistical power analysis and sample size calculation
interpret_p() Standalone p-value interpreter

Usage

One command does it all 

library(statease)

# Descriptive statistics
analyze(x = c(23, 45, 12, 67, 34), var_name = "Exam Scores")

# Independent samples t-test (auto-detected), with a study design note
analyze(x = c(23,45,12,67,34), y = c(19,38,22,51,29),
        var_name = "Scores",
        context = "convenience sample, not randomly assigned")

# Check assumptions before deciding on a test
analyze(x = c(23,45,12,67,34), y = c(19,38,22,51,29),
        check = TRUE)

# Non-parametric alternative (auto-detected)
analyze(x = c(23,45,12,67,34), y = c(19,38,22,51,29),
        nonparam = TRUE, var_name = "Scores")

# Correlation (auto-detected)
analyze(x = c(23,45,12,67,34), y = c(19,38,22,51,29),
        var1_name = "Exam Score", var2_name = "Study Hours")

# Chi-square (auto-detected)
analyze(
  x = c("Yes","No","Yes","Yes","No"),
  y = c("Male","Female","Male","Female","Male")
)

# One-way ANOVA (auto-detected)
df <- data.frame(
  score = c(23,45,12,67,34,89,56,43,78,90,11,34),
  group = rep(c("A","B","C"), each = 4)
)
analyze(formula = score ~ group, data = df)

# Two-way ANOVA (auto-detected)
df2 <- data.frame(
  score  = c(23,45,12,67,34,89,56,43,78,90,11,34),
  method = rep(c("Online","Traditional"), each = 6),
  gender = rep(c("Male","Female"), times = 6)
)
analyze(formula = score ~ method * gender, data = df2)

# Simple linear regression (auto-detected)
df3 <- data.frame(
  exam_score  = c(23,45,12,67,34,89,56,43,78,90),
  study_hours = c(2,5,1,7,3,9,6,4,8,10)
)
analyze(formula = exam_score ~ study_hours, data = df3)

# Power analysis
analyze(test_type = "ttest.two", effect_size = 0.5)

# Interpret any p-value
interpret_p(0.03, context = "treatment vs control group")

What an assumption check actually looks like

Every relevant _interpret() function prints its assumption checks automatically, whether or not anything is wrong:

  Assumption Checks:
    Normality (Group 1)    : PASSED   (Shapiro-Wilk p = 0.342)
    Normality (Group 2)    : WARNING  (Shapiro-Wilk p = 0.012, may not be normal)
    Equal variances        : PASSED   (Levene's p = 0.501)

  NOTE: Assumption checks are diagnostic tools and may be
  influenced by sample size and other characteristics of the
  data. Passing a check does not prove that an assumption is
  satisfied, and a warning does not automatically invalidate
  the analysis. Interpret these results alongside your
  knowledge of the data.

Checks are labelled one of three ways: - PASSED : the package tested this and found no evidence of a problem - WARNING : the package detected something worth your attention - NOTE : something relevant to interpretation that the package cannot test automatically (independence of observations, for example, is a property of how the data was collected, not something computable from the numbers themselves)

Why statease?

Most R output gives you numbers. statease gives you numbers, a plain-English interpretation, and by default, the assumption context needed to read that interpretation responsibly. It’s built for: - Students learning statistics - Researchers who want fast, readable output without skipping diagnostics - Educators teaching statistical concepts

Changelog

v1.4.0

  • Assumption checks are now printed by default in every relevant _interpret() function, rather than requiring a separate call to check_assumptions()
  • Added a context argument across all inferential functions and analyze(), letting users describe their study design and have it echoed back alongside the interpretation
  • Added a numerical separation diagnostic to logistic_interpret()
  • reg_interpret() and mlr_interpret() now check homoscedasticity and residual independence in addition to normality; mlr_interpret() also checks multicollinearity (VIF)
  • check_assumptions()’s regression logic now shares its diagnostic calculations with reg_interpret() and mlr_interpret(), rather than three separate implementations
  • Fixed a boundary bug in power_interpret() where an effect size exactly equal to a Cohen’s convention threshold was labelled one category too high
  • Fixed a bug where anova2_interpret()’s printed report did not display the Sum of Squares type
  • Fixed a bug where chisq_interpret() triggered R’s internal chi-squared approximation warning twice
  • Fixed a bug where lm()/glm() fitted inside a wrapper function could cause car::ncvTest() to fail silently when computing homoscedasticity
  • Several formatting fixes in non-parametric test output

v1.3.0

v1.2.1

  • Fixed non-parametric interpretation — stochastic superiority correctly reported instead of median differences

v1.2.0

v1.1.0

v1.0.0

License

MIT