Skip to contents

Visualize categorical metadata effects on numeric values.

Source: R/boxplot.r
stats_boxplot.Rd

Visualize categorical metadata effects on numeric values.

Usage

stats_boxplot(
  df,
  x = NULL,
  y = attr(df, "response"),
  layers = "x",
  stat.by = x,
  facet.by = NULL,
  colors = TRUE,
  shapes = TRUE,
  patterns = FALSE,
  test = "auto",
  flip = FALSE,
  stripe = NULL,
  ci = "ci",
  level = 0.95,
  p.adj = "fdr",
  p.top = Inf,
  outliers = NULL,
  xlab.angle = "auto",
  p.label = 0.05,
  strata = NULL,
  caption = TRUE,
  ...
)

Arguments

df

The dataset (data.frame or tibble object). "Dataset fields" mentioned below should match column names in df. Required.

x

A categorical metadata column name to use for the x-axis. Or NULL, which groups all samples into a single category.

y

A numeric metadata column name to use for the y-axis. Default: attr(df, 'response')

layers

The elements to draw for each group, such as boxes, violins, dots, or error bars. One or more of c("bar", "box" ("x"), "violin", "dot", "strip", "crossbar", "errorbar", "linerange", "pointrange"). Single letter abbreviations are also accepted. For instance, c("box", "dot") is equivalent to c("x", "d") and "xd". Default: "x"

stat.by

Dataset field with the statistical groups. Must be categorical. Default: NULL

facet.by

Dataset field(s) to use for faceting. Must be categorical. Default: NULL

colors

How to color the groups. Options are:

TRUE -

Automatically select colorblind-friendly colors.

FALSE or NULL -

Don't use colors.

a palette name -

Auto-select colors from this set. E.g. "okabe"

character vector -

Custom colors to use. E.g. c("red", "#00FF00")

named character vector -

Explicit mapping. E.g. c(Male = "blue", Female = "red")

See "Aesthetics" section below for additional information. Default: TRUE

shapes

Shapes for each group. Options are similar to colors's: TRUE, FALSE, NULL, shape names (typically integers 0 - 17), or a named vector mapping groups to specific shape names. See "Aesthetics" section below for additional information. Default: TRUE

patterns

Patterns for each group. Options are similar to colors's: TRUE, FALSE, NULL, pattern names ("brick", "chevron", "fish", "grid", etc), or a named vector mapping groups to specific pattern names. See "Aesthetics" section below for additional information. Default: FALSE

test

Method for computing p-values. 'auto' runs pairwise Mann-Whitney U tests ('wilcox') when x and stat.by are the same field (or 'emmeans' when strata is given), and Kruskal-Wallis tests ('kruskal') across the stat.by groups within each x category otherwise. Beta diversity boxplots use PERMANOVA ('adonis2'), 'mrpp', or 'betadisper' on the samples instead. 'none' skips the statistics. See stats_table() and bdiv_stats(). Default: 'auto'

flip

Transpose the axes, so that taxa are present as rows instead of columns. Default: FALSE

stripe

Shade every other x position. Default: same as flip

ci

How to calculate min/max of the crossbar, errorbar, linerange, and pointrange layers. Options are: "ci" (confidence interval), "range", "sd" (standard deviation), "se" (standard error), and "mad" (median absolute deviation). The center mark of crossbar and pointrange represents the mean, except for "mad" in which case it represents the median. Intervals are drawn as computed, including below zero, unless the y-axis is transformed: bounds below zero are drawn at zero on a "sqrt" or "log1p" axis, and bounds at or below zero are drawn at the smallest positive y value on a log axis. Default: "ci"

level

The confidence level for calculating a confidence interval. Default: 0.95

p.adj

Method to use for multiple comparisons adjustment of p-values. Run p.adjust.methods for a list of available options. The adjustment spans every p-value in the returned table (all metrics, strata, pairs, and positions). Default: "fdr"

p.top

Only display taxa with the most significant differences in abundance. If p.top is >= 1, then the p.top most significant taxa are displayed. If p.top is less than one, all taxa with an adjusted p-value <= p.top are displayed. Recommended to be used in combination with the taxa parameter to set a lower bound on the mean abundance of considered taxa. Default: Inf

outliers

Show boxplot outliers? TRUE to always show. FALSE to always hide. NULL to only hide them when overlaying a dot or strip chart. Default: NULL

xlab.angle

Angle of the labels at the bottom of the plot. Options are "auto", '0', '30', and '90'. Default: "auto".

p.label

Minimum adjusted p-value to display on the plot with a bracket.

p.label = 0.05 -

Show p-values that are <= 0.05.

p.label = 0 -

Don't show any p-values on the plot.

p.label = 1 -

Show all p-values on the plot.

If a numeric vector with more than one value is provided, they will be used as breaks for asterisk notation. Default: 0.05

strata

Dataset field with the blocking/subject variable. Permutation tests permute within strata; regression models include a random intercept per stratum. Default: NULL

caption

Add methodology caption beneath the plot. Default: TRUE

...

Additional parameters to pass along to ggplot2 functions. Prefix a parameter name with a layer name to pass it to only that layer. For instance, d.size = 2 ensures only the points on the dot layer have their size set to 2. Unprefixed parameters go to every layer whose function accepts them (e.g. scales = "free" for the facets). An argument that no layer accepts is an error.

Value

A ggplot2 plot. The computed data points, ggplot2 command, stats table, and stats table commands are available as $data, $code, $stats, and $stats$code, respectively.

Aesthetics

The default colors are colorblind-friendly for up to 12 groups: Okabe and Ito's palette ("okabe") for up to 8, and Paul Tol's ("tol") for 9-12. Of the other palettes, "bright", "vibrant", "muted", "light", and "carto" are also designed for color vision deficiencies. The available categorical palette names are: "okabe", "carto", "r4", "polychrome", "tol", "bright", "light", "muted", "vibrant", "tableau", "classic", "alphabet", "tableau20", "kelly", and "fishy".

Patterns are added using the fillpattern R package, which must be installed to use patterns. Options are "brick", "chevron", "fish", "grid", "herringbone", "hexagon", "octagon", "rain", "saw", "shingle", "rshingle", "stripe", and "wave", optionally abbreviated and/or suffixed with modifiers. For example, "hex10_sm" for the hexagon pattern rotated 10 degrees and shrunk by 2x. See fillpattern::fill_pattern() for complete documentation of options.

Shapes can be given as per base R - numbers 0 through 17 for various shapes, or the decimal value of an ascii character, e.g. a-z = 65:90; A-Z = 97:122 to use letters instead of shapes on the plot. Character strings may used as well.

The dot and strip layers use the ggbeeswarm package when it is installed, and jittered points otherwise.

See also

Other visualization: adiv_boxplot(), adiv_corrplot(), bdiv_boxplot(), bdiv_corrplot(), bdiv_heatmap(), bdiv_ord_plot(), plot_heatmap(), rare_corrplot(), rare_multiplot(), rare_stacked(), stats_corrplot(), taxa_boxplot(), taxa_corrplot(), taxa_forest(), taxa_heatmap(), taxa_stacked(), theme_rbiom()

Examples

    library(rbiom)

    df <- adiv_table(rarefy(hmp50))
    stats_boxplot(df, x = "Body Site")

    stats_boxplot(df, x = "Sex", stat.by = "Body Site", layers = "be")