# Introduction to carpenter

#### 2019-01-31

Note: At present, carpenter only creates tables that you would typically see as the ‘basic characteristics’ or ‘descriptive statistics’ table in most biomedical articles. The plan is to include other common table structures, but they haven’t been implemented yet. If you have any suggestions, let me know as an Issue!

In biomedical research, there are certain types of tables that are often included in the article. For instance, some basic statistics between the treatment and control group. Or maybe it is between males and females, before and after an intervention, and so on. Often these tables are a hassle to create and are prone to needing updates based on slight changes in the data or from reviewer comments. carpenter was designed to make creating these tables easily and for them to be easily updated when data changes.

## Minor caveat/consideration

Before going over the code, I should mention a key note and assumption: how the table is eventually presented is determined by how the data looks when it is passed into the carpenter functions. A good example is ‘before and after’ studies, where researchers may store data where each row is the participant and the remaining columns being the before and after values of a measure (e.g. Weight_before and Weight_after as two columns). In this case, the data should ideally be formatted more like this:

ID Visit Weight
1 Before 50
1 After 54
2 Before 65
2 After 70

See how the data is stored in ‘long’ format. This is the form of data that carpenter was designed to deal with. For more details on what is tidy data see resources at the end.

## Code usage

The underlying design principal for carpenter is that you create an outline of what the table should look like before finally creating the table. This is how carpenters also work: they sketch what the product will look like before actually starting to build anything. There are four ‘outlining’ functions, several carpenter statistics functions, and one final ‘building’ function:

• outline_table() starts the outlining process.
• add_rows() adds variables to the row or rows with associated descriptive statistics.
• stat_*() type statistic functions to be used with add_rows() (a list of available statistics can be found using ?carpenter::table_stats).
• renaming() for customizing the naming of the rows and table headers.
• build-table() for finally building the table (uses the pander package).

These functions are chained together using the wonderful magrittr %>% pipe. If you’ve never used this package or the pipe, take a look at the vignette on introducing it. So, let’s do some coding:

library(magrittr)
library(carpenter)
head(iris)
#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#> 1          5.1         3.5          1.4         0.2  setosa
#> 2          4.9         3.0          1.4         0.2  setosa
#> 3          4.7         3.2          1.3         0.2  setosa
#> 4          4.6         3.1          1.5         0.2  setosa
#> 5          5.0         3.6          1.4         0.2  setosa
#> 6          5.4         3.9          1.7         0.4  setosa
outline_table(iris, 'Species') 
#> # A tibble: 0 x 0

You’ll notice that the outline_table function returned a tibble of 0 rows and 0 columns. That’s because we haven’t added anything else to the table! carpenter is waiting for more information. So we add rows by:

outline_table(iris, 'Species') %>%
add_rows('Sepal.Length', stat_meanSD) 
#> # A tibble: 1 x 4
#>   Variables    setosa    versicolor virginica
#>   <chr>        <chr>     <chr>      <chr>
#> 1 Sepal.Length 5.0 (0.4) 5.9 (0.5)  6.6 (0.6)

You see it has now added a row to the table. Adding more rows:

outline_table(iris, 'Species') %>%
add_rows('Sepal.Width', stat_medianIQR) 
#> # A tibble: 3 x 4
#>   Variables    setosa        versicolor    virginica
#>   <chr>        <chr>         <chr>         <chr>
#> 1 Sepal.Length 5.0 (0.4)     5.9 (0.5)     6.6 (0.6)
#> 2 Petal.Length 1.5 (0.2)     4.3 (0.5)     5.6 (0.6)
#> 3 Sepal.Width  3.4 (3.2-3.7) 2.8 (2.5-3.0) 3.0 (2.8-3.2)

See how this can make creating these tables very easy. Let’s remove the . from the row names and fix up the table header names:

tab <- outline_table(iris, 'Species') %>%
renaming('header', c('Measures', 'Setosa', 'Versicolor', 'Virginica')) %>%
renaming('rows', function(x) gsub('\\.', ' ', x))
tab
#> # A tibble: 3 x 4
#>   Measures     Setosa        Versicolor    Virginica
#>   <chr>        <chr>         <chr>         <chr>
#> 1 Sepal Length 5.0 (0.4)     5.9 (0.5)     6.6 (0.6)
#> 2 Petal Length 1.5 (0.2)     4.3 (0.5)     5.6 (0.6)
#> 3 Sepal Width  3.4 (3.2-3.7) 2.8 (2.5-3.0) 3.0 (2.8-3.2)

And finally build the table into a Markdown format for easy insertion into R Markdown documents:

build_table(tab)
Measures Setosa Versicolor Virginica
Sepal Length 5.0 (0.4) 5.9 (0.5) 6.6 (0.6)
Petal Length 1.5 (0.2) 4.3 (0.5) 5.6 (0.6)
Sepal Width 3.4 (3.2-3.7) 2.8 (2.5-3.0) 3.0 (2.8-3.2)

If you have factor/discrete data, you can even use include these variables:

library(dplyr, quietly = TRUE)
#>
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>
#>     filter, lag
#> The following objects are masked from 'package:base':
#>
#>     intersect, setdiff, setequal, union
mtcars %>%
mutate(
gear = as.factor(gear),
vs = as.factor(vs)
) %>%
outline_table('vs') %>%
renaming('header', c('Measures', 'V-engine', 'Straight engine')) %>%
renaming('rows', function(x) x %>%
gsub('drat', 'Read axle ratio', .) %>%
gsub('mpg', 'Miles/gallon', .) %>%
gsub('gear', 'Number of gears', .)) %>%
build_table()
Measures V-engine Straight engine
Miles/gallon 16.6 (3.9) 24.6 (5.4)
Read axle ratio 3.2 (3.1-3.7) 3.9 (3.7-4.1)
Number of gears
- 3 12 (66.7%) 3 (21.4%)
- 4 2 (11.1%) 10 (71.4%)
- 5 4 (22.2%) 1 (7.1%)

Pretty easy eh?

Sometimes, though, you don’t need to compare multiple columns, but instead need to only show one column. Easy, just don’t include a header in the outline_table()!

iris %>%
outline_table() %>%
build_table() 
Measures Values
Sepal.Length 5.8 (0.8)
Sepal.Width 3.0 (2.8-3.3)

If you don’t use R Markdown, you can continue the chain into write.csv(). Use the finish = FALSE argument to build_table() to prevent the Markdown table from being created.!

iris %>%
outline_table() %>%
write.csv('table1.csv', row.names = FALSE)