Formatting, printing and exporting tables"

In insight: Easy Access to Model Information for Various Model Objects

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

if (!requireNamespace("poorman", quietly = TRUE) ||
  !requireNamespace("gt", quietly = TRUE)) {
  knitr::opts_chunk$set(eval = FALSE)
} else {
  library(poorman)
  library(gt)
}

The difference between a dataframe and its render

Most of objects encountered throughout the {easystats} packages are "tables", i.e., a 2D matrix with columns and rows. In R, these objects are often, at their core, data frames. Let's create one to use as an example:

library(insight)

df <- data.frame(
  Variable = c(1, 3, 5, 3, 1),
  Group = c("A", "A", "A", "B", "B"),
  CI = c(0.95, 0.95, 0.95, 0.95, 0.95),
  CI_low = c(3.35, 2.425, 6.213, 12.1, 1.23),
  CI_high = c(4.23, 5.31, 7.123, 13.5, 3.61),
  p = c(0.001, 0.0456, 0.45, 0.0042, 0.34)
)

df

When I display in in the console (calling an object - e.g. df - is actually equivalent to calling print(df)), the output looks alright, but it could be improved. Some packages, such as {knitr}, have functions to create a nicer output. For instance, in markdown, so that it can be nicely rendered in markdown documents when copied:

knitr::kable(df, format = "markdown")

| Variable|Group |   CI| CI_low| CI_high|      p|
|--------:|:-----|----:|------:|-------:|------:|
|        1|A     | 0.95|  3.350|   4.230| 0.0010|
|        3|A     | 0.95|  2.425|   5.310| 0.0456|
|        5|A     | 0.95|  6.213|   7.123| 0.4500|
|        3|B     | 0.95| 12.100|  13.500| 0.0042|
|        1|B     | 0.95|  1.230|   3.610| 0.3400|

Or HTML, which again makes it look great in HTML files (such as this webpage you're reading):

knitr::kable(df, format = "html")

The insight workflow

The {insight} package also contains function to improve the "printing", or rendering, of tables. Its design dissociates two separate and independent steps: formatting and exporting.

Formatting

The purpose of formatting is to improve a given table, while still keeping it as a regular R data frame, so that it can be for instance further modified by the user.

format_table(df)

As you can see, format_table() modifies columns, turning number into characters (so that it has the same amount of digits), and detecting confidence intervals. This is usually combined with column-specific formatting functions, like format_p():

df %>%
  mutate(p = format_p(p, stars = TRUE)) %>%
  format_table()

Using unicode symbols as effect size names

With use_symbols = TRUE, it is possible to render certain effect size names as symbols, if these are used as column names. Note that this only works on OS X or Linux, or on Windows from R 4.2 or higher.

x <- data.frame(
  phi_adjusted = 0.3,
  Glass_delta = 0.4,
  Epsilon2 = 0.7,
  R2 = 0.4
)

# standard output
format_table(x)

# column names of effect sizes as symbols
format_table(x, use_symbols = TRUE)

In combination with export_table() (see next section), this will give you nicely formatted tables.

export_table(format_table(x, use_symbols = TRUE))

Exporting

The next step is exporting, which takes a data frame and renders it in a given format, so that it looks good in the console, or in markdown, HTML or latex.

export_table(df)

For markdown or HTML, simply change the format argument to markdown ("md")...

export_table(df, format = "md")

...or HTML format.

export_table(df, format = "html")

This can be combined with format_table().

df %>%
  format_table(ci_brackets = c("(", ")")) %>%
  export_table(format = "html")

TODO: What about display?

insight documentation built on Nov. 26, 2023, 5:08 p.m.