7 Introduction to the tidyverse

The tidyverse is a collection of packages by the creators of RStudio that share an approach to data science.

The authors model data science like this:

Data science life cycle

The tidyverse packages replace some of the base R functions with alternatives that are intended to be more user friendly for data scientists who are following this life cycle.

We will only be covering a few of the packages from the tidyverse.

library(tidyr)
library(dplyr)
## 
## Attaching package: 'dplyr'
## The following object is masked from 'package:kableExtra':
## 
##     group_rows
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(magrittr)
## 
## Attaching package: 'magrittr'
## The following object is masked from 'package:tidyr':
## 
##     extract

7.1 Defining tidy data

In “tidy” data, every column is a variable, every row is an observation, and every cell contains a single observation. Is the birthweight data frame tidy? Why or why not?

7.2 Pipes: combining tidyverse functions

The tidyverse employs piping to send the output of one function to another function, rather than the nesting used in base r. The “pipe” is written with a greater than symbol sandwiched between two percent signs, like this: %>%.

experiment %>%
  filter(low.birthweight == TRUE) %>%
  select(birth.date, length, birthweight, smoker)
Table 7.1: Tidyverse filter and select.
birth.date length birthweight smoker
7/18/1968 46 2.05 TRUE
9/16/1968 48 1.92 TRUE
1/8/1968 47 2.66 TRUE
3/23/1967 50 2.51 TRUE
9/27/1968 43 2.65 FALSE
4/2/1968 48 2.37 TRUE 
# equivalent to:
experiment[experiment$low.birthweight == TRUE, c("birth.date", "length", "birthweight", "smoker")]
Table 7.2: Base R subsetting.
birth.date length birthweight smoker
5 7/18/1968 46 2.05 TRUE
8 9/16/1968 48 1.92 TRUE
10 1/8/1968 47 2.66 TRUE
13 3/23/1967 50 2.51 TRUE
33 9/27/1968 43 2.65 FALSE
35 4/2/1968 48 2.37 TRUE

7.3 Transforming data

The separate() function makes the conversion of the “birth.date” column into “month,” “day,” and “year” trivial.

experiment %>%
  filter(low.birthweight == TRUE) %>%
  select(birth.date, length, birthweight, smoker) %>%
  separate(col = birth.date, sep = "[/]", into = c("month", "day", "year"))
Table 7.3: Tidyverse separate.
month day year length birthweight smoker
7 18 1968 46 2.05 TRUE
9 16 1968 48 1.92 TRUE
1 8 1968 47 2.66 TRUE
3 23 1967 50 2.51 TRUE
9 27 1968 43 2.65 FALSE
4 2 1968 48 2.37 TRUE

The mutate() function adds a new column based on data contained in the existing columns.

experiment %>%
  filter(low.birthweight == TRUE) %>%
  select(birth.date, length, birthweight, smoker) %>%
  mutate(d = birthweight / length)
Table 7.4: Tidyverse mutate.
birth.date length birthweight smoker d
7/18/1968 46 2.05 TRUE 0.0445652
9/16/1968 48 1.92 TRUE 0.0400000
1/8/1968 47 2.66 TRUE 0.0565957
3/23/1967 50 2.51 TRUE 0.0502000
9/27/1968 43 2.65 FALSE 0.0616279
4/2/1968 48 2.37 TRUE 0.0493750

7.4 Summarizing data

The group_by() and summarize() functions apply a function to a group defined by one or more categorical variables.

experiment %>%
  group_by(smoker) %>%
  summarize(mean.birthweight = mean(birthweight))
Table 7.5: Tidyverse summarize grouped by smoker.
smoker mean.birthweight
FALSE 3.509500
TRUE 3.134091 
experiment %>%
  group_by(smoker, low.birthweight) %>%
  summarize(mean.birthweight = mean(birthweight))
## `summarise()` has grouped output by 'smoker'. You can override using the `.groups` argument.
Table 7.6: Tidyverse summarize grouped by smoker and birthweight.
smoker low.birthweight mean.birthweight
FALSE FALSE 3.554737
FALSE TRUE 2.650000
TRUE FALSE 3.378824
TRUE TRUE 2.302000

To change the order of rows, use arrange(). To return one or more specified rows, use slice().

experiment %>%
  group_by(smoker) %>%
  select(smoker, birthweight, length, head.circumference, weeks.gestation) %>%
  slice_max(order_by = birthweight, n = 5)
Table 7.7: Tidyverse slice, ordered on birthweight.
smoker birthweight length head.circumference weeks.gestation
FALSE 4.55 56 34 44
FALSE 4.32 53 36 40
FALSE 4.10 58 39 41
FALSE 4.07 53 38 44
FALSE 3.94 54 37 42
TRUE 4.57 58 39 41
TRUE 3.87 50 33 45
TRUE 3.86 52 36 39
TRUE 3.64 53 38 40
TRUE 3.59 53 34 40

The pivot_longer() and pivot_wider() functions rearrange data, decreasing or increasing the number of columns. The use of this will become more evident during visualization.

experiment %>%
  filter(low.birthweight == TRUE) %>%
  select(smoker, `miR-16`, `miR-21`, `miR-146a`, `miR-182`) %>%
  pivot_longer(cols = c(`miR-16`, `miR-21`, `miR-146a`, `miR-182`),
               names_to = "gene",
               values_to = "expression")
Table 7.8: Tidyverse pivot_longer.
smoker gene expression
TRUE miR-16 60
TRUE miR-21 48
TRUE miR-146a 89
TRUE miR-182 65
TRUE miR-16 70
TRUE miR-21 51
TRUE miR-146a 84
TRUE miR-182 48
TRUE miR-16 61
TRUE miR-21 51
TRUE miR-146a 97
TRUE miR-182 59
TRUE miR-16 47
TRUE miR-21 40
TRUE miR-146a 87
TRUE miR-182 63
FALSE miR-16 64
FALSE miR-21 47
FALSE miR-146a 104
FALSE miR-182 42
TRUE miR-16 61
TRUE miR-21 47
TRUE miR-146a 77
TRUE miR-182 60

7.5 Exercise 4: converting between base R and Tidyverse

Reproduce the table 7.5 or table 7.6 using base R. Use Tidyverse functions to answer the question you addressed in exercise 3.