Complex data processing with dplyr

The plyr function has started a small revolution within the R community, due to its fantastic functions for data manipulation/wrangling. Its creators released an improved package, called dplyr, which is essentially a plyr on steroids: some of its machinery has been rebuilt via the Rcpp package (this allows the inclusion of C++ functions in R) achieving a much better performance.

The dplyr function is tightly connected to the maggrittr package, which allows us to use the %>% notation to chain operations one after the other. For instance, we could do something like sort %>% filter %>% summarize, wherein the output of each operation will be sent as an input for the next one. This becomes really powerful when writing a lot of code for data manipulation.

Getting ready

The dplyr package can be installed via install.packages() as usual.

How to do it...

We will use it to do data wrangling with the very famous mtcars dataset. This dataset contains several attributes (both numeric and categorical) for several car models. We will combine several of the dplyr functions to compute the mean horsepower and mean miles per gallon for each gear type. We will finally sort the result by the mean horsepower.

1. Import the dplyr library:

library(dplyr)

2. The main step is grouping the data by gear; we calculate the mean horsepower, and the sum of the miles per gallon by gear. We then destroy the group structure and we sort the results by mean horsepower using the following code block:

mtcars %>% group_by(am,gear) %>% summarise(mean_hp = mean(hp),sum_mpg = sum(mpg)) %>% ungroup %>% arrange(mean_hp)

How it works...

The beauty of the dplyr package is that it allows us to chain several operations together one after the other. This greatly reduces the complexity of our code, generating very simple and readable code. The ungroup statement is not strictly needed, but it destroys the grouped structure that we have had up to that point. In certain situations, this is necessary, because we might need to work on the ungrouped data after some grouping or summarizing is done.

There's more...

The dplyr function can be used to do dozens of interesting things, as it allows us to summarize, sort, group, compute ranks, arrange accumulators by group, and so on.

See also

The dplyr function's official website (https://dplyr.tidyverse.org) is an excellent reference source for learning it.