Challenge 9 Solutions

Creating a function

Challenge Overview

Today’s challenge is simple. Create a function, and use it to perform a data analysis / cleaning / visualization task:

Examples of such functions are: 1) A function that reads in and cleans a dataset.
2) A function that computes summary statistics (e.g., computes the z score for a variable).
3) A function that plots a histogram.

That’s it!

A function that reads in and cleans a dataset

This very simple function that reads in the SNL Cast Dataset and converts the first and last episode date columns to date format¹.

We also include an optional argument that allows us to only look at those cast members who were an anchor on Weekend Update. This can simplify the process of filtering later on.

read_snl_casts <- function(f,update_only=F){
  d <- read_csv(f) %>%
    mutate(across(c(first_epid,last_epid),~ymd(as.character(.x))))
  if(update_only){
    d <- d %>%
      filter(update_anchor)
  }
  return(d)
}
snl_casts <- here("posts","_data","snl_casts.csv") %>%
  read_snl_casts()
snl_casts

snl_update_casts <- here("posts","_data","snl_casts.csv") %>%
  read_snl_casts(update_only = T)
snl_update_casts

A function that computes summary statistics

To use such a function, we’ll first import the hotel bookings dataset, which includes plenty of quantitative variables.

hotels <- here("posts","_data","hotel_bookings.csv") %>%
  read_csv()
hotels

Our function will compute the mean, median, and standard deviation for a variable. Specifically, the function will take a vector (or a column from a data frame) and output a tibble of these summary statistics.

sum_stat <- function(x){
  stat <- tibble(
    mean=mean(x,na.rm=T),
    median=median(x,na.rm=T),
    sd=sd(x,na.rm=T)
  )
  return(stat)
}
sum_stat(hotels$lead_time)

map_dfr(list(hotels$lead_time,hotels$babies,hotels$days_in_waiting_list),sum_stat) %>%
  mutate(id=c("lead time","babies","days in waiting list")) %>%
  relocate(id,.before=everything())

A function that plots a histogram

This is also a very simple function - essentially a wrapper function around geom_histogram() that plots a histogram of a variable. We’ll use the hotel bookings dataset again - specifically the lead_time variable.

We can use a double-brace around the name of the column we want a histogram of. For more details around this approach (known as data masking), visit this link.

make_my_hist <- function(dat, colname, fill="purple", xlab="x", ylab="n"){
  dat %>%
    ggplot(aes({{colname}}))+
    geom_histogram(fill=fill)+
    labs(x=xlab,
         y=ylab)
}
make_my_hist(hotels,lead_time,xlab="lead time")

make_my_hist(hotels,lead_time,fill="blue")

We can save our plot as an object and add to it using the + operator.

my_graph <- make_my_hist(hotels,lead_time,xlab="lead time")
my_graph_w_title <- my_graph +
  labs(title="Lead time histogram")
my_graph_w_title

Conclusion

There’s clearly a lot we can do with functions. This solution comprises a small chunk of how to do so. A crucial aspect of programming with functions is knowing the extent to which your function will work. For example, the read_snl_casts() function will fail on any dataset that isn’t snl_casts.csv, while the make_my_hist() function will work on many different datasets.

Functions are great for automating processes and avoiding copy-paste. We will see more of how to do so in challenge 10 - purrr.

Footnotes

1. Note that most values here are NA