groupdiff: Test of group differences
In Rkabacoff/qacr: Functions to Facilitate Data Science Projects

Description Usage Arguments Details Value See Also Examples

One-way analysis (ANOVA or Kruskal-Wallis Test) with post-hoc comparisons and plots

groupdiff(
  formula,
  data,
  method = c("anova", "kw"),
  digits = 2,
  horizontal = FALSE,
  posthoc = TRUE
)

`formula`	use the form `Y ~ X` where `Y` is the response variable, and `X` is the categorical explanatory variable. `X` will be coerced to be a factor.
`data`	a data frame.
`method`	character. Either `"anova"`, or `"kw"` (see details).
`digits`	Number of significant digits to print.
`horizontal`	logical. If `TRUE`, boxplots are plotted horizontally.
`posthoc`	logical. If `TRUE`, the default, perform pairwise post-hoc comparisons (TukeyHSD for ANOVA and Conover Test for Kuskal Wallis). This test will only be performed if there are 3 or more levels for X.

The groupdiff function performs one of two analyses:

anova: A one-way analysis of variance, with TukeyHSD post-hoc comparisons.
kw: A Kruskal Wallis Rank Sum Test, with Conover Test post-hoc comparisons.

In each case, summary statistics and a grouped boxplots are provided. In the parametric case, the statistics are n, mean, and standard deviation. In the nonparametric case the statistics are n, median, and median absolute deviation. If posthoc = TRUE, pairwise comparisons of superimposed on the boxplots. Groups that share a letter are not significantly different (p < .05), controlling for multiple comparisons.

a list with 3 components:

result: omnibus test
summarystats: summary statistics
plot: ggplot2 graph

kwAllPairsConoverTest, multcompLetters.

# parametric analysis
groupdiff(hp ~ gear, mtcars)

# nonparametric analysis
groupdiff(popularity ~ vehicle_style, cardata,
          method="kw", horizontal=TRUE)