Contingency Tables

library(apa7)
library(dplyr)
library(tidyr)
library(flextable)
library(ftExtra)

Data

d <- mtcars |>
  select(Gears = gear, Transmission = am) |>
  mutate(Transmission = factor(Transmission, labels = c("Automatic", "Manual")))

The apa_chisq contingency table function

The default output of the apa_chisq function is a contingency table with a chi-square test of independence in the table note. The output is a flextable, which can be modified with flextable commands.

apa_chisq(d)

Transmission	3		4		5
	n	%	n	%	n	%
Automatic	15	100.0%	4	33.3%	0	0.0%
Manual	0	0.0%	8	66.7%	5	100.0%
Note. χ2 (2) = 20.94, p < .001, Adj. Cramer’s V = .78

Styling Options

The table can be styled in many ways:

apa_chisq(
  d,
  font_size = 16,
  line_spacing = 3,
  text_color = "darkred",
  border_width = 2,
  border_color = "gray",
  font_family = "Arial"
)

Transmission	3		4		5
	n	%	n	%	n	%
Automatic	15	100.0%	4	33.3%	0	0.0%
Manual	0	0.0%	8	66.7%	5	100.0%
Note. χ2 (2) = 20.94, p < .001, Adj. Cramer’s V = .78

# No note
apa_chisq(d, note = NA)

Transmission	3		4		5
	n	%	n	%	n	%
Automatic	15	100.0%	4	33.3%	0	0.0%
Manual	0	0.0%	8	66.7%	5	100.0%

# Custom note
apa_chisq(
  d,
  note = "This is a *custom* note written in **markdown** $x > \\omega$."
)

Transmission	3		4		5
	n	%	n	%	n	%
Automatic	15	100.0%	4	33.3%	0	0.0%
Manual	0	0.0%	8	66.7%	5	100.0%
Note. This is a custom note written in markdown x > ω.

What if I want something completely different?

There are a lot of options out there. The flextable package has the proc_freq function, which has the ability to include row, column, and total percentages in the table.

mtcars %>%
  proc_freq(row = "gear", col = "vs") %>%
  theme_apa()

gear		vs
		0	1	Total
3	Count	12 (37.5%)	3 (9.4%)	15 (46.9%)
	Mar. pct (1)	66.7% ; 80.0%	21.4% ; 20.0%
4	Count	2 (6.2%)	10 (31.2%)	12 (37.5%)
	Mar. pct	11.1% ; 16.7%	71.4% ; 83.3%
5	Count	4 (12.5%)	1 (3.1%)	5 (15.6%)
	Mar. pct	22.2% ; 80.0%	7.1% ; 20.0%
Total	Count	18 (56.2%)	14 (43.8%)	32 (100.0%)
(1) Columns and rows percentages

These can be turned off selectively:

mtcars %>%
  proc_freq(
    row = "gear",
    col = "vs",
    include.row_percent = FALSE,
    include.column_percent = FALSE,
    include.table_percent = FALSE
  ) %>%
  apa_style(table_width = .5)

gear	vs
	0	1	Total
3	12	3	15
4	2	10	12
5	4	1	5
Total	18	14	32

The flextable package’s tabulator function has considerable power in creating a wide variety of descriptive tables. In general, the tabulator function requires that you calculate the statistics first, and then you specify where they should go.

Here I calculate the means, standard deviations, and sample sizes within each cell of a contingency table.

d_tension <- warpbreaks |>
  summarise(
    m = mean(breaks, na.rm = TRUE),
    sd = sd(breaks, na.rm = TRUE),
    n = n(),
    .by = c(wool, tension)
  ) %>%
  rename(Wool = wool) |>
  mutate(
    tension = factor(
      tension,
      labels = paste(c("Low", "Medium", "High"), "Tension")
    )) 
  
d_tension |> 
  flextable::tabulator(
    rows = c("Wool"),
    columns = "tension",
    M = as_paragraph(m),
    SD = as_paragraph(sd),
    N = as_paragraph(n)
  ) |> 
  flextable::as_flextable() |>
  italic(i = 2, part = "header") |>
  align(j = 1, align = "center") |> 
  theme_apa()

Wool		Low Tension				Medium Tension				High Tension
		M	SD	N		M	SD	N		M	SD	N
A		44.6	18.1	9		24.0	8.7	9		24.6	10.3	9
B		28.2	9.9	9		28.8	9.4	9		18.8	4.9	9

Processing Data Yourself

What if you want it to look a little different? Unfortunately, you might have to do some of the heavy lifting yourself. The pivot_wider_name_first function is a wrapper around tidyr::pivot_wider that prints and sorts column names as “name_variable” instead of “variable_name”. I wanted a shortcut because I frequently needed to make a shortcut for setting names_glue = "{tension}_{.value}" and names_vary = "slowest" when names_from = "tension".

A subtle difference in this version of the table is that the columns are decimal aligned, which is evident in the standard deviation column.

Also, whereas previously the “stub header” (Wool column) was middle aligned, it is now top aligned, which is consistent with comparable examples in the APA Style Manual. I am not sure if this is a real rule in APA style, and I actually prefer middle alignment for situations like this.

d_tension |> 
  rename(M = m, SD = sd) |> 
  pivot_wider_name_first(
    values_from = c(M, SD, n),
    names_from = tension
  ) |>
  apa_flextable(column_formats = column_formats(accuracy = .1)) |>
  align(j = 1, align = "center") 

Wool	Low Tension				Medium Tension				High Tension
	M	SD	n		M	SD	n		M	SD	n
A	44.6	18.1	9		24.0	8.7	9		24.6	10.3	9
B	28.2	9.9	9		28.8	9.4	9		18.8	4.9	9