Into the tidyverse: dplyr
I completed an introductory course in using R for statistical analysis. The instructions were in the underlying base functions in R (base R).
This article compares the code for data manipulation, using dplyr and base R. This article only covers basic tasks.
Data management and manipulation
Introductory courses ask students to undertake simple manipulate of data frames. These courses may use the base R functions.
Since 2016, there have been a collection of packages which share the same design philosophy. As the packages focus on ‘tidy’ data, designers call this the tidyverse.
In this article, I will use the standard iris data-set in R.
One basic step is to count how many rows meet one criterion. We want to know how many plants in our data-set are in the Setosa species.
In base R, we turn the data frame into a series of ‘TRUE’ and ‘FALSE’ entries. When we sum, that turns ‘TRUE’ values into 1, and ‘FALSE’ into 0. The sum then counts the rows which meet the criterion.
sum(iris$Species == "setosa")
In dplyr, we often use the ‘forward-pipe’ operator
%>%. This operator will forward a value or resulting expression into the next expression. That way, we can build our query — step by step.
filter(Species == "setosa") %>%
Instead of a single count, we can produce frequency tables. In base R, that uses the length function on the Species column.
tapply(iris$Species, iris$Species, length)
In the tidyverse, we count by species:
iris %>% count(Species)
Another common manipulation is to create a subset. This is similar in base R and tidyverse code.
iris_setosa_df <- subset(iris, iris$Species == "setosa")iris_setosa_df <- iris %>% filter(Species == "setosa")
We may wish to take the top rows of the data set. In base R, the order function puts the vector elements in a certain order. That order is then used on the data set. The head function takes the top six rows:
head(iris_setosa_df[order(iris_setosa_df$Petal.Length, decreasing = TRUE), ], n = 6)
In tidyverse, we use arrange and desc to arrange the rows in descending order (as the names suggest). The slice_head function to take the first six rows:
slice_head(n = 6)
We often want to create new variables. There are two different ways of doing this in base R:
iris_setosa_df$sepal_ratio <- iris_setosa_df$Sepal.Length / iris_setosa_df$Sepal.Widthiris_setosa_df[["sepal_ratio"]] <- iris_setosa_df[["Sepal.Length"]] / iris_setosa_df[["Sepal.Width"]]
For this purpose, we can use the mutate function in the tidyverse:
iris_setosa_df <- iris_setosa_df %>%
mutate(sepal_ratio = Sepal.Length / Sepal.Width)
The tidyverse functions may offer a plainer way to learn for people new to R.