The tidyverse makes heavy use of the R concept of forward pipes. Forward pipes, represented via %>%, are provided by the magrittr package, which should be automatically loaded by the tidyverse.

(if not, you can load it via load.packages("magrittr"))

A forward pipe forwards the variable on the left into the first argument to the function on the right, e.g.

"kitten" %>% print()

will forward the string “kitten” so that it is the first argument to the function print. Hence this is exactly identical to

print("kitten")

and so

[1] "kitten"

is printed.

You may ask why this is useful? It is useful because it enables you to chain together a lot of functions. For example, the tidyverse dply package provides the function filter, for filtering data.

So;

cats %>% filter(Sex=="F")
# A tibble: 47 x 3
Sex   BodyWeight HeartWeight
<chr>      <dbl>       <dbl>
1 F            2           7
2 F            2           7.4
3 F            2           9.5
4 F            2.1         7.2
5 F            2.1         7.3
6 F            2.1         7.6
7 F            2.1         8.1
8 F            2.1         8.2
9 F            2.1         8.3
10 F            2.1         8.5
# … with 37 more rows

has filtered the cats data set from the last page to return a tibble that contains data only for female cats.

This was identical to typing;

filter(cats, Sex=="F")

The power comes that we can now chain filters, e.g.

cats %>% filter(Sex=="F") %>% filter(BodyWeight > 2.5)
# A tibble: 11 x 3
Sex   BodyWeight HeartWeight
<chr>      <dbl>       <dbl>
1 F            2.6         8.7
2 F            2.6        10.1
3 F            2.6        10.1
4 F            2.7         8.5
5 F            2.7        10.2
6 F            2.7        10.8
7 F            2.9         9.9
8 F            2.9        10.1
9 F            2.9        10.1
10 F            3          10.6
11 F            3          13  

We can then use the dplyr summarise function to call calculate_mean on a specified row of this filtered data, e.g.

cats %>%
filter(Sex=="F") %>%
filter(BodyWeight>2.5) %>%
summarise(mean=calculate_mean(HeartWeight))

will output

# A tibble: 1 x 1
mean
<dbl>
1  10.2

as the mean average of the heart weight in grams of female cats whose body weight is greater than 2.5 kg.

Note how we have split this over multiple lines, putting the forward pipe %>% at the end so that it is clear that the line continues (use Shift+Enter to start a new line without running the command in the R Console).

To save this to a variable, we would use the assign <- as normal, hence the full code should be;

average_heart_weight <- cats %>%
filter(Sex=="F") %>%
filter(BodyWeight>2.5) %>%
summarise(mean=calculate_mean(HeartWeight))

Yes, this is a very dense bit of code. This is typical for R. You will often see very dense blocks of code that use forward pipes to push data through several functions, resulting in a final output result. As you can see, it is important that you name your variables, data, columns and functions clearly, so that it is easier for future readers of your code to understand what is going on.

Finally, note that average_heart_weight is a 1x1 tibble. You can extract the actual numeric value by typing as.numeric(average_heart_weight).

EXERCISE

Calculate the average heart weight of male cats whose body weight is greater than or equal to 3.0 kg.

Look back at the vignette you found when searching for the Pearson’s product-moment correlation. How much more of this vignette do you now understand? Have a go at installing packages that you don’t recognise, and using ? in RStudio to get help on the functions that are new to you.