Dancing between character vectors and formula objects

I was determined to exactly reproduce a composite box-and-whisker plot that I had been seeing in the book Introduction to Statistical Learning. The data are from a set of 1,250 observations called Smarket that is found in the R package that goes with the book, known as ISLR. It’s a pretty simple plot:

First, after loading the package with library(ISLR), one may want to take a cursory look at the data frame, which has 9 variable columns, using View(head(Smarket)).

I was able to successfully put together the plot with this code:

ylabel <- "Percentage change in S&P"
xlabel <- "Today's Direction"
valnames <- c("Down", "Up")
hue <- c("blue", "red")

layout(matrix(c(1, 2, 3), nrow = 1, ncol = 3, byrow = TRUE))
boxplot(Lag1 ~ Direction, data = df,
        ylab = ylabel, xlab = xlabel,
        names = valnames,
        col = hue,
        main = "Yesterday")

boxplot(Lag2 ~ Direction, data = df,
        ylab = ylabel, xlab = xlabel,
        names = valnames,
        col = hue,
        main = "Two Days Previous")

boxplot(Lag3 ~ Direction, data = df,
        ylab = ylabel, xlab = xlabel,
        names = valnames,
        col = hue,
        main = "Three Days Previous")
layout(1)

I realised, however, that I was seriously violating the DRY principle, so I tried to come up with a function instead. I struggled a bit with this because I didn’t know how to supply a character vector argument and place it into the “formula” bit that is required as the first argument of the version of the boxplot() function I had used i.e. the y ~ x part.

After scouring the documentation a bit – I tried out as.name(), messed around with sQuote() and dQuote(), all to no avail – I discovered the documentation behind ?formula, and BAMMM!, I got it.

This is how I put it to get the exact same plot:

box.stock <- function(dat, column, tt) {
    par(mfrow = c(1, 3))
    for (i in 1:3) {
        boxplot(as.formula(paste(column[i], "~ Direction")),
                data = dat,
                ylab = "Percentage change in S&P",
                xlab = "Today's Direction",
                names = c("Down", "Up"), 
                col = c("blue", "red"),
                main = tt[i])
    }
}

choice <- c("Lag1", "Lag2", "Lag3")
title <- c("Yesterday", "Two Days Previous", "Three Days Previous")

box.stock(df, choice, title)

Now how’s that?

This experience helped me to better appreciate the value of using functions in general. Of course, with a little tweaking, the above code could be used to draw many more plots, indeed  as much as is within available computing or application resources.