Skip to contents

Co-Missingness Correlation Matrix

Source: R/plot_comissingness_matrix.R
plot_comissingness_matrix.Rd

Heatmap of the correlation between the missingness indicators of columns (1 = missing, 0 = observed). Values near 1 mean two columns are missing together (an instrument or a line that failed and took several measurements with it); values near 0 are consistent with unrelated missingness. With `cluster = TRUE` the columns are ordered by hierarchical clustering of `1 - correlation`.

Usage

plot_comissingness_matrix(
  data,
  vars = NULL,
  cluster = TRUE,
  k = NULL,
  palette = "Reds",
  label = NULL
)

Arguments

data

A data frame.

vars

Columns to use (default all).

cluster

Order the columns by hierarchical clustering (default `TRUE`).

k

Optional number of clusters to cut the tree into; adds a `cluster` column to the plot data.

palette

A sequential ColorBrewer palette name (default `"Reds"`); the scale runs 0 to 1 (negative correlations are drawn as 0).

label

Print the correlation on the tiles (default `TRUE` up to 15 columns).

Value

A ggplot; its `data` holds `var1`, `var2`, `r` (and `cluster` for `var1` when `k` is given).

Details

Columns with no missing value, or only missing values, have no variation in their indicator and are left out (listed in the attribute `"dropped"` of the result).

See also

[missmap()], [features_percent_miss()]

Examples

d <- data.frame(a = c(1, NA, 3, NA, 5, 6), b = c(1, NA, 3, NA, 5, NA), c = c(NA, 2, 3, 4, 5, 6))
plot_comissingness_matrix(d)