The R Programming Language

R is an open source implementation of the S and S-Plus statistical programming languages. R is free to download and use, as is the R Studio IDE.

The base R language has several functions for plotting. In addition, there are a few very useful libraries for data visualization. We will use ggplot2, which was created by Hadley Wickham using the concept of the “grammar of graphics,” a concept first articulated by statistician Leland Wilkinson.

Many of the visualizations that follow center around the mtcars data set. mtcars is part of the data sets that come with base R. It is very frequently referred to in R tutorials. The help file describes it like this:

The data was extracted from the 1974 Motor Trend US magazine, and comprises fuel consumption and 10 aspects of automobile design and performance for 32 automobiles (1973–74 models).

The first five rows of the data set are presented below.

                   mpg cyl disp  hp drat    wt  qsec vs am gear carb
Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1    4    4
Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1    4    4
Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1    4    1
Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0    3    1
Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0    3    2

We are interested in the variable mpg (miles per gallon).

Particularly, we want to know which of the other variables in our data set might be influential in determining (or predicting) miles per gallon. In a realistic research scenario, we would use multiple regression analysis. Here we will restrain ourselves to making some simple comparisons of mpg with the other variables in the data set, one at a time.

The goal is to see if there are any strong patterns that emerge that might guide us in asking more sophisticated questions.

The following 5 slides examine how the distribution of miles per gallon is influenced by the factor variables in the data set. The factor variables are:

• Number of cylinders
• Transmission type (auto vs. manual)
• Engine type
• Number of gears
• Number of carburetors

The next 5 slides are scatter plots of mpg against the continuous variables in the data set, treated as explanatory variables. These plots allow us to think in terms of lines of least squares, and to visually inspect the data for any weird outliers or obvious patterns that would influence the outcome of a linear regression.

The other five continuous variables in the data set are:

• Displacement (in cubic inches)
• Horsepower
• Rear axle ration
• Weight
• Quarter mile time

The rest of the slides will rely heavily on the ggplot2 library for their data visualizaitons. We will start by using the quick plotting command qplot() to continue our look at the mtcars data set. Afterwards, we will look at a variety of data with the more customizable ggplot() command.

We can call the library like this:

library(ggplot2)

We can ask a lot of questions about the data before formalizing our hypothesis and conducting more rigourous statistics. With our mtcars data we have been asking about the differing breakdowns of mpg over various subsets. We could also ask about the distributions of the other variables. By simply plotting one single-variable distribution after another, we could check to see which, if any, potential explanatory variables have skewed distributions or bimodal plots. We can look for outliers as well as missing values. In larger, more complicated data sets, we could use feature detection algorithms or other ideas from the field of machine learning.

Let's say we're interested in the wt (weight) variable, as it appears to be influential upon mpg. We're going to focus on the relationship between those two variables, especially as it is itself influenced by the presence of other variables.

Again, in practice we would want to use multiple regression as a supplement to, or outright replacement for, this approach. For the time being, this will give us a chance to see the qplot() function in action.

The next slides will show a variety of uses for the ggplot() command. The code and the resulting graphs are taken from the excellent book The R Graphics Cookbook, by Winston Chang. The figure numbers correspond to the figures in this text.

In addition to the ggplot2 library, we will need the gcookbook library that accompanies the Chang text. We will also need the plyr library very briefly.

library(gcookbook)
library(plyr)

ggplot2 uses “geoms” to support the ggplot() function. Geoms add functionality to the graphs, allowing them to understand how we want to display the data. We will add geom_bar() to our ggplot() function calls for the next few slides.

This graph takes a subset of 2001's best 25 hitters in baseball, sorts them by league, and then reorders the data before presenting it as a faceted dot plot.

tophit <- tophitters2001[1:25, ] # top 25 hitters
nameorder <- tophit$name[order(tophit$lg, tophit$avg)]
tophit$name <- factor(tophit$name, levels=nameorder)
ggplot(tophit, aes(x=avg, y=name)) +
  geom_segment(aes(yend=name), xend=0, color="grey50") +
  geom_point(size=3, aes(color=lg)) +
  scale_color_brewer(palette="Set1", limits=c("NL", "AL"), guide=FALSE) +
  theme_bw() +
  theme(panel.grid.major.y = element_blank()) +
  facet_grid(lg ~ ., scales="free_y", space="free_y")

Next we will see how ggplot2 handles line graphs. We add geom_line() and geom_point() at various spots.

sunspotyear <- data.frame (
  Year = as.numeric(time(sunspot.year)),
  Sunspots = as.numeric(sunspot.year)
)

ggplot(sunspotyear, aes(x=Year, y=Sunspots)) +
  geom_area(fill="blue", alpha=0.2) +
  geom_line()

Next, we'll see how ggplot2 handles scatter plots.

Here, I set up the function to trim the extra ID column and add a season column. I will use this function on each file as I read the file into the workspace.

edit_tables <- function(df, year) {
  df <- df[, 2:28]
  season <- rep(toString(year), nrow(df))
  df <- cbind(season, df)
}

Now I can use that function to edit each data frame right when it is created.

b2012 <- read.csv('blazers2012.csv')
b2012 <- edit_tables(b2012, 2012)

b2013 <- read.csv('blazers2013.csv')
b2013 <- edit_tables(b2013, 2013)

b2014 <- read.csv('blazers2014.csv')
b2014 <- edit_tables(b2014, 2014)

b2015 <- read.csv('blazers2015.csv')
b2015 <- edit_tables(b2015, 2015)

b2016 <- read.csv('blazers2016.csv')
b2016 <- edit_tables(b2016, 2016)

# Now I need all that data in one "object."
# I can make one data frame out of many by using
# R's "rbind" (or row-bind) function.
blz <- rbind (b2012, b2013, b2014, b2015, b2016)

Now our project has a data frame called “blz”. That will be the main item of interest. We could delete the other data frames to preserve memory, but these aren't very large files and it doesn't make sense to worry about them right now.

The current blz data frame doesn't have very descriptive column names. In some case they are long and unwieldy for the purposes of display in a data frame.

Here I define some shorter names. Some aren't very descriptive, but for the purpose of translating NBA season averages they should make sense.

my_cols <- c('seas', 'pname', 'age',
             'games', 'start', 'mins',
             'fg', 'fga', 'fgp',
             't3p', 't3pa', 't3pp',
             't2p', 't2pa', 't2pp', 'efgp',
             'ft', 'fta', 'ftp',
             'orb', 'drb', 'trb',
             'ast', 'stl', 'blk',
             'tov', 'pf', 'pts')

Here I assign those column names to the blz data frame.

colnames(blz) <- my_cols

And the data frame is ready.

I want to create a data frame of players who played more than one season in the time frame under consideration.

Here's the basic functionality. Running this will PRINT the different subsets, but it does not STORE them anywhere. This is here for demonstration purposes so we can se what we're getting.

for (i in 1:length(player)) {
  if (nrow(player[[i]]) > 1) {
    print (player[[i]])
  }
}

mult_seasons <- function(mylist) {
  collector <- data.frame()
  for (i in 1:length(mylist)) {
    if (nrow(mylist[[i]]) > 1) {
      collector <- rbind(collector, mylist[[i]])
    }
  }
  return(collector)
}

As an example, we can create “twoplus”, a data frame of season averages of players with at least 2 years on the team during they years in question.

twoplus <- mult_seasons(player)
head(twoplus)

Here are graphs of the distributions of player ages and rebounding averages over the time period in question.

Here are lists of players by length of tenure with the team over that time period. We'll get the data graphed, and then make some fine tuning adjustments to get it in order.

playersums <- table(blz$pname)
playerdf <- as.data.frame(playersums)
colnames(playerdf) <- c('pname', 'count')
playerdf <- transform(playerdf,
                      pname=reorder(pname, count))