Text map of where the missing values are

Draws the missingness of a whole table as a grid, one character per cell block – a console counterpart of visdat::vis_miss() that needs no graphics device, so it works over SSH, in a log, and inside a capsule's plain-text summary.

Rows are binned so the map fits height lines; a block is drawn at the shade its missing proportion falls in. Seeing the table at once is the point: a diagonal band, a block of rows, or a single ragged column are all instantly recognisable shapes that a column of percentages is not.

Usage

missgrid(data, height = 20L, width = 12L)

Arguments

data

A data frame whose missing cells are drawn as a grid.

height

Whole number, default 20: the maximum number of lines (row blocks) in the map.

width

Whole number, default 12: column names are cut to this many characters in the vertical legend.

Value

A character vector of the map's lines, invisibly; the map is printed with cat() as a side effect.

Details

The rows are split into at most height consecutive blocks of nearly equal size; each map line is one block, labelled with its first row number, and each map column is one column of data. A cell's shade is its block's proportion missing rounded to the nearest quarter: blank (0), light, medium, dark, full (1). Where the console cannot show Unicode the shades are " .-+#". Column names are written vertically above the map, cut to width characters. A table with no rows prints "(no rows)". Errors: data not a data frame or without columns; height not a positive whole number.

See also

misspat(), missruns()

Examples

set.seed(1)
df <- data.frame(
  complete = 1:100,
  block = c(rep(NA, 30), 31:100),
  scattered = ifelse(stats::runif(100) < 0.3, NA, 1),
  mostlygone = c(1:10, rep(NA, 90))
)
missgrid(df)
#> ── Missingness map ───────────────────────────────────────────────
#>          cbsm
#>          olco
#>          moas
#>          pctt
#>          lktl
#>          eey
#>          trg
#>          eeo
#>          dn
#>          e
#>        1  █▒ 
#>        6  █░ 
#>       11  █▒█
#>       16  █ █
#>       21  █▒█
#>       26  █░█
#>       31   ░█
#>       36   ░█
#>       41    █
#>       46   ░█
#>       51   ▒█
#>       56   ░█
#>       61   ░█
#>       66   ▒█
#>       71    █
#>       76    █
#>       81    █
#>       86   ▓█
#>       91   ▒█
#>       96    █
#> 
#>   legend: ' ' none  '█' all missing
#> ──────────────────────────────────────────────────────────────────
#> 

# a coarser map with short labels
missgrid(df, height = 5, width = 4)
#> ── Missingness map ───────────────────────────────────────────────
#>          cbsm
#>          olco
#>          moas
#>          pctt
#>        1  █░▒
#>       21  ▒░█
#>       41   ░█
#>       61   ░█
#>       81   ░█
#> 
#>   legend: ' ' none  '█' all missing
#> ──────────────────────────────────────────────────────────────────
#> 

try(missgrid(df, height = 0))
#> Error : `height` must be a single positive integer