Describe Relationship: Numeric x by Categorical g

Computes, prints and plots descriptive statistics for a numeric variable x grouped by a categorical variable g. The function is dispatched automatically by desc(y ~ g, data) when y is numeric and g categorical.

Usage

.descNQ(x, g, ...)

# S3 method for class 'Desc.nq'
print(x, digits = NULL, ...)

# S3 method for class 'Desc.nq'
plot(x, main = x$meta$main, which = 1, ...)

Arguments

x

a numeric variable for .descNQ(), or an object of class "Desc.nq" for the print and plot methods

digits

currently unused

...

further arguments. In .descNQ() unused, in print() passed to bedrock::printCharMatrix(). In plot() they are passed on to the plot function selected by which (see section Plots, where each function is linked) and go unchanged to every selected plot.

main

main title for the plot; defaults to the title stored in x$meta$main

which

integer vector selecting the plots to draw, one plot per element, see section Plots. Default 1.

g

a categorical grouping variable (factor or coercible to factor)

Value

.descNQ() returns an object of class c("Desc.nq", "Desc") with components:

tab

group-wise summary table

test

result of the Kruskal-Wallis test

vtest

result of Levene's test

eta

effect size

The plot method returns x invisibly.

Details

The function summarizes the distribution of x across levels of g and performs nonparametric tests of group differences.

Computed statistics

• Group-wise descriptive statistics (mean, median, SD, IQR, counts)

• Kruskal-Wallis test

• Effect size (\(\eta^2\)) based on the Kruskal-Wallis statistic

• Levene's test for homogeneity of variance

Interpretation The Kruskal-Wallis test evaluates whether the distribution of x differs between groups defined by g. The effect size \(\eta^2\) provides a standardized measure of group differences.

Plots

plot() draws one of three displays of x by group, selected by which. All arguments in ... go straight to the underlying function; its help page lists what can be set.

which = 1

Boxplots by group, drawn by pharos::plotBox(). Axes are labelled with the variable names.

which = 2

Overlaid kernel density estimates, one per group, drawn by pharos::plotDens().

which = 3

Density and boxplot combined per group, drawn by pharos::plotDensBox().

main defaults to the title stored in the object and is passed to every plot.

See also

desc(), desc.qn(), desc.nn(), desc.qq(), kruskal.test(), lumen::leveneTest()

Plot functions: pharos::plotBox(), pharos::plotDens(), pharos::plotDensBox()

Other desc: desc(), desc.Date(), desc.factor(), desc.nn, desc.numeric(), desc.qn, desc.qq, desc.table(), desc.ts()

Examples

# basic usage via desc()
desc(temperature ~ area, Pizza)
#> ────────────────────────────────────────────────────────────────────────────── 
#> temperature ~ area (Pizza) (Desc.nq)
#> 
#> Summary:
#> pairs: 1209, valid: 1161 (96.0%), missings: 48 (4.0%), groups: 3
#> 
#>           Brent   Camden   Westminster
#> mean     51.139   47.420        44.258
#> median   53.400   50.300        45.900
#> sd        8.734   10.111         9.836
#> iqr      10.500   12.200        13.200
#> n           467      335           359
#> np        40.2%    28.9%         30.9%
#> NAs           7        9            22
#> zeros         0        0             0
#> 
#> Kruskal-Wallis rank sum test:
#>   Kruskal-Wallis chi-squared = 115.83, df = 2, p-value < 2.2e-16
#>   η² = 0.098 (moderate)
#> 
#> Levene's Test for Homogeneity of Variance (center = median):
#>   F = 5.3473, num df = 2, denom df = 1158, p-value = 0.004879
#> 
#> Warning message:
#>   Grouping variable contains 10 NAs (0.8%).
#> 


# store result, print and plot separately
d <- desc(temperature ~ area, Pizza, plotit = FALSE)
d
#> ────────────────────────────────────────────────────────────────────────────── 
#> temperature ~ area (Pizza) (Desc.nq)
#> 
#> Summary:
#> pairs: 1209, valid: 1161 (96.0%), missings: 48 (4.0%), groups: 3
#> 
#>           Brent   Camden   Westminster
#> mean     51.139   47.420        44.258
#> median   53.400   50.300        45.900
#> sd        8.734   10.111         9.836
#> iqr      10.500   12.200        13.200
#> n           467      335           359
#> np        40.2%    28.9%         30.9%
#> NAs           7        9            22
#> zeros         0        0             0
#> 
#> Kruskal-Wallis rank sum test:
#>   Kruskal-Wallis chi-squared = 115.83, df = 2, p-value < 2.2e-16
#>   η² = 0.098 (moderate)
#> 
#> Levene's Test for Homogeneity of Variance (center = median):
#>   F = 5.3473, num df = 2, denom df = 1158, p-value = 0.004879
#> 
#> Warning message:
#>   Grouping variable contains 10 NAs (0.8%).
#> 

# the three plots
plot(d, which = 1)                     # boxplots          -> plotBox()

plot(d, which = 2)                     # densities         -> plotDens()

plot(d, which = 3)                     # density + boxplot -> plotDensBox()


# pipe
desc(temperature ~ area, Pizza) |> plot(which = 2)