Lab 3.2: Tidyverse: Pipes and Dplyr

This week’s agenda: learning to master pipes and dplyr.

# Load the tidyverse!
library(tidyverse)

Pipes to base R

For each of the following code blocks, which are written with pipes, write equivalent code in base R (to do the same thing).

• 1a.

letters %>%
  toupper %>%
  paste(collapse="+") 

## [1] "A+B+C+D+E+F+G+H+I+J+K+L+M+N+O+P+Q+R+S+T+U+V+W+X+Y+Z"

• 1b.

"     Ceci n'est pas une pipe     " %>% 
  gsub("une", "un", .) %>%
  trimws

## [1] "Ceci n'est pas un pipe"

• 1c. (Still use ggplot)

rnorm(1000) %>% 
  data.frame(x = .) %>%
  ggplot(.) +  # I'm giving you a hint here
    geom_histogram(aes(x = x, y = ..density..)) + 
    labs(title = "N(0,1) draws")

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

• 1d.

rnorm(1000) %>% 
  hist(breaks=30, plot=FALSE) %>% # use the ?hist to figure out what this does when plot=FALSE
  .[["density"]] %>%
  max

## [1] 0.405

Base R to pipes

For each of the following code blocks, which are written in base R, write equivalent code with pipes (to do the same thing).

• 2a. Hint: you’ll have to use the dot ., as seen above in Q1b, or in the lecture notes.

paste("Your grade is", sample(c("A","B","C","D","R"), size = 1))

## [1] "Your grade is D"

• 2b. Hint: you can use the dot . again, in order to index state.name directly in the last pipe command.

state.name[which.max(state.x77[,"Illiteracy"])] 

## [1] "Louisiana"

• 2c. Hint: if x is a list of length 1, then x[[1]] is the same as unlist(x).

str_url <- 
  paste0("https://raw.githubusercontent.com/benjaminleroy/",
         "36-350-summer-data/master/Week1/endgame.txt")

# Base R:
lines <- readLines(str_url)
text <- paste(lines, collapse = " ")
words <- strsplit(text, split = "[[:space:]]|[[:punct:]]")[[1]]
wordtab <- table(words)
wordtab <- sort(wordtab, decreasing = TRUE)
head(wordtab, 10)

## words
##         the    to     I     a   and   you     s    of    it 
## 10146   780   553   478   466   408   375   374   282   261

• 2d. Hint: the only difference between this and the last part is the line words = words[words != ""]. This is a bit tricky line to do with pipes: use the dot ., once more, and manipulate it as if were a variable name.

# Base R:
lines <- readLines(str_url)
text <- paste(lines, collapse = " ")
words <- strsplit(text, split = "[[:space:]]|[[:punct:]]")[[1]]
words <- words[words != ""]
wordtab <- table(words)
wordtab <- sort(wordtab, decreasing = TRUE)
head(wordtab, 10)

## words
## the  to   I   a and you   s  of  it  in 
## 780 553 478 466 408 375 374 282 261 251

Shark attack data, revisited

Below we read in the similar data.frame shark_attacks containing information about victims of shark attacks we’ve seen in previous labs. (Note the difference of location - I changed it a little bit.)

shark_attacks <- read.csv("https://raw.githubusercontent.com/benjaminleroy/36-350-summer-data/master/Week2/shark-attacks-clean.csv", stringsAsFactors = TRUE)

• 3a. Write and document a function time_factor_to_numeric that transforms a time (storaged as a stringed factor) into a numerical value. Additionally, make sure that if the value is NA, the function returns NA. Use the below my_fac and my_fac2 as an example of what format is expected to be inputed. Also uncomment the code below to check you function. Reminder: use the Roxygen2 shortcut to help document your function.

time_factor_to_numeric <- function(time_fac) {
  # fill in and document
  NULL
}

my_fac <- factor(c("13h30"))
my_fac2 <- factor(c(NA))

# my_fac_numeric <- time_factor_to_numeric(my_fac)
# my_fac_numeric == 13.5
# my_fac2_numeric <- time_factor_to_numeric(my_fac2)
# is.na(my_fac2_numeric)

• 3b. Attempt your function on the factor vector my_fac3 found below (Use the commented out code below to do so). If it doesn’t work continue to 3c. after defining time_factor_to_numeric_no_vec <- time_factor_to_numeric as seen below. If your function does vectorize (i.e. passes the all.equal statement), define time_factor_to_numeric_no_vec as another function with the same structure as time_factor_to_numeric but use strsplit and see that it doesn’t vectorize.

my_fac3 <- factor(c("05h15", NA, "23h59"))

#my_fac3_numeric <- time_factor_to_numeric(my_fac3)
#all.equal(my_fac3_numeric, c(5.25, NA, 23 + 59/60))

time_factor_to_numeric_no_vec <- time_factor_to_numeric

• 3c. Write a new function time_factor_to_numeric_vec that takes in a vector and returns a vector (use the same test as in 3b). For this implimentation use an sapply and your time_factor_to_numeric_no_vec. Document as well.

• 3d. In R, as you’ve seen with indexing and more, we approach of problems through a vectorized framework. Below you’ll see that we define a new function time_factor_to_numeric_vectorized that wraps the function Vectorize around your non-vectorized function. Use the same test as seen in part 3b and 3c to see if this function works well

time_factor_to_numeric_vectorized <- Vectorize(time_factor_to_numeric_no_vec)

dplyr attacks sharks

With your vectorized function we’ll start exploring the dplyr tools we saw in lecture.

• 4a. Quick clean up. Convert the Time column in the shark_attacks to numeric using time_factor_to_numeric_vectorized. Hint: use mutate_at(), and reassign shark_attacks to be the output.

• 4b. Using dplyr’s mutate function create 3 new numeric columns year, month, and day for the shark_attacks data.frame Before you do, check the class of the Date class using the dplyr pipes and the pull function.

• 4c. Suppose you’re writing a story about shark attacks in the United States that were fatal. You’re interested in tracking down family of some of these individuals. Moreover, for journalistic purposes - it’s probably best that you select families based on how recently the attack happened. First, create a new data.frame fatal_shark_attacks_usa that contain information of all attacks in the US that were fatal.

• 4d. Using fatal_shark_attacks_usa produce a table that contains individuals that meet the rest of the conditions above, and show the top 3 individuals’s Name, Age, Area of attack, and year of attack. Hint: this will use arrange, select, and slice. I expect this will take some thinking - maybe pull out a piece of paper and write down the goals and try to decide how they relate to dplyr functions.

• 4e. Remembering that ggplot does some tabulation internally in certain geoms: Use dplyr piping and functions to explore the following question. Are there particular months of the year where its there are fewer fatal attacks than others (all countries)?

• 4f. That seems weird. Facet the above graphic on country. Describe what you see. What is the month with the most fatal attacks in the US?

• 4g. Challenge: Use the built in data vector month.abb to create a new column in shark_attacks month_abb that is a factor variable will the correct month associated with the attack and is ordered. Update 3f and 3e figures with this information. Also add scale_x_discrete(drop = FALSE) to your plots. What does this do?