Guide to R Commands

read_csv

read_csv is used for loading comma-separated-format (CSV) files. This is a popular format for transmitting data, and works well with spreadsheet programs like Excel. Excel can read CSV files and can also export/save to CSV format.

In RStudio you can use the Import Dataset > From CSV menu option, and it will create the appropriate read_csv command for you.

# A local file:
myData <- read_csv("myDatafile.csv")
# A file from the internet
myData <- read_csv("http://hanoverstatslabs.github.io/resources/datasets/driving.csv")

data

data is used to load data that was saved in an R package. It is typically followed by the View command.

If used with no arguments, as data(), it will list all datasets available in all loaded packages.

# See all available datasets
data()
# Load a dataset with a specific name
data(brfss)
# Load a dataset with a specific name
data(brfss, package="hanoverbase")
# Follow it with the `View` command
View(brfss)

View

View is used to bring up a spreadsheet view of a dataset. The dataset must have been previously created, by read_csv, data or some other method.

# First load a dataset
load(brfss)
# Then view it
View(brfss)

head, tail

head and tail can be used in order to extract the first few, or the last few, cases from a dataset. They can often be used as the last step in a piping sequence in order to see the top results from our operation.

Both functions take two arguments: The name of the object (often provided via piping) and the number of cases to show, which defaults to around 6.

# Show only the first few cases from the dataset
data(brfss)
head(brfss)
# Same thing using piping
brfss %>% head()
# Showing 10 cases
head(brfss, 10)
brfss %>% head(10)
# State populations, top 6
data(counties)
sum(~pop2010|state, data=counties) %>% sort() %>% tail()

filter

TODO

The Formula Interface

Most of the commands we are using expect the variables to be specified via a formula interface. A formula in general looks as follows:

<lhs>~<rhs>|<groups>

Here the left-hand-side may be completely absent and the vertical line and groups part may also be absent. Both sides contain the variables to be considered, and multiple variables can be separated by a plus sign.

lhs

The variables on the left-hand-side are the target variables and we are interested in how they are influenced by the other variables. When doing a scatterplot, lhs corresponds to the y axis.

rhs

The variables on the right-hand-side are the source variables and we are interested in themselves if there is no left-hand-side, or how they influence the variables in the left-hand-side if there are any.

groups

The variables in the groups section are meant to be variables used to break down the results. If used in a graph situation for example, the graph would have a different panel for each value of the group variables.

# Looking at general health on its own
~genhealth
# Compare physical health against mental health,
# break down by general health category
physhealth~menthealth|genhealth
# Look at both physical health and mental health
~physhealth+menthhealth

Piping

The piping operatior, %>%, is a very useful operator that allows us to simplify the writing of complex steps. The operator takes the value on its left side, and uses it as the first argument in the function on its right side. Here is one example:

healthVsExercise <- tally(~genhealth|exerciseany, data=brfss)
healthVsExercise %>% t() %>% barchart()

The first line creates a new table, healthVsExercise, with the tallies. The second line takes that table, and feeds it into the t function, and then feeds the result into the barchart function. That line is exactly the same as the following much harder to read expression:

barchart(t(healthVsExercise))

summary

summary can be used on the whole dataset, to get a quick overview of the various variables and their summaries.

summary(brfss, digits=3, maxsum=6)

tally

tally is used to produce frequency tables and cross-tabulation tables for categorical variables. The first argument should be a formula involving the variables to use, and it should be followed by a data=<dataset> option specifying which dataset to use.

# There is a higher percent of Very Good/Excellent health people amongst those
# that exercise vs those that do not.
tally(~genhealth|exerciseany, data=brfss, format="percent", useNA="no")

Aggregating Functions

There is a long list of commands that all follow a similar syntax, and all aggregate values of some variable, possibly broken down by other categorical variables. They all have the syntax:

<command>(<formula>, ...options)

The formula has the form <aggregatingVariable>~<breakDown1>+<breakDown2>... where the variable on the LHS is the scalar variable to be aggregated, and the variables in the RHS are the categorical variables by which to break the groups down.

The usual commands are:

• mean for the variable’s mean
• median for the variable’s median
• sd for the standard deviation
• var for the variance (square of standard deviation)
• iqr for the interquartile range
• favstats for the five-number summary, mean, median and count
• range for the range of the values (maximum minus minimum)
• max for the maximum of the values
• min for the minimum of the values
• sum for the sum of the values
• prod for the product of the values

# Stats on number of days per month with physical/mental health problems,
# broken down by general health
mean(physhealth~genhealth, data=brfss, na.rm=TRUE)
mean(menthealth~genhealth, data=brfss, na.rm=TRUE)
favstats(menthealth~genhealth, data=brfss, na.rm=TRUE)
# Total state population by adding the county populations
sum(pop2010~state, data=counties, na.rm=TRUE)
# Stats for percent of blacks in county, broken down by state
favstats(black~state, data=counties, na.rm=TRUE)

Common Graph Options

These options can be added to most graph commands:

• xlab="..." specifies the label for the x axis.

• ylab="..." specifies the label for the y axis.

• main="..." specifies the main graph title, which appears at the top.

• col=... specifies what colors to use for the various elements. Look at the Color Specifications section for more details.

• pch=... can be used on graphs with points to choose what point-style to use. The options are summarized in this image:

  

  The icons 21-25 can use a fill-in (background) color, via the bg=... command. Use c(..., ...) for multiple values.

• lty=... can be used on graphs with lines to choose the line style to use. The graph above shows the options for line types. Use c(..., ...) for multiple values.

• lwd=... can be used on graphs with lines to specify the thickness of the lines, 1 being the standard size. For instance lwd=1.4 will draw the lines 40% thicker.

• cex=... can be used to change the size of the objects in the graph, 1 being the standard size. For instance cex=1.2 will draw the objects 20% bigger.

• xlim=c(a, b) sets the range on the x axis to be from a to b.

• ylim=c(a, b) sets the range on the y axis to be from a to b.

• auto.key and key: These options are used for to add legends to graphs that need them. For details look at the corresponding section.

• layout: TODO

Legend Specifications

Legends are needed when we use a grouping variable, to identify the different colors and symbols used. Adding a default legend is easy to do. However, customizing the legend and the graph in general becomes trickier.

To specify the legend you have two options you can add to graphs:

auto.key

The auto.key option is a convenient way to add a legend, but it will only work if you don’t set custom colors to your graph. Its form is auto.key=list(...) where the dotted area may be empty or contain any of the following options: - space="..." will specify which side the legend goes in (top, bottom, left, right). The default is top. - columns=... will specify in how many columns to split the legend. The default is one column, so all values form a single column.

key

The key option allows for more fine-tuned legends, but it is also more complex to write. This is definitely an advanced feature. With this option you have to manually specify everything that goes into the graph, including the various labels for the grouping variable. It would typically be constructed in steps, like in this example:

# First we specify the colors we will use. We use 5 keys here.
myColors <- brewer.pal(5, "RdPu")
# Next we create the levels
myLabels = list(levels(brfss$genhealth))
# Now we create a list of key components. We use "text" to set the labels,
# "columns" to specify how many columns to show, and "rectangles" to
# tell the legend to draw colored rectangles with our colors in.
myKey = list(text=myLabels, columns=4, rectangles=list(col=myColors))
# Finally, we use the key in a graph We must specify the colors again
healthVsExercise <- tally(~genhealth|exerciseany, data=brfss, format="percent", useNA="no")
healthVsExercise %>% t() %>% barchart(col=myColors, key=myKey)

Instead of, or in addition to, rectangles, we could also use lines and points to add those symbols. You can use most of the common graph options to specify these lines and symbols, just like we used col above.

Color Specifications

Graph colors can be specified in a variety of different ways. We outline these different ways here. They are to be used with the col=... graph option.

By name

R has over 650 built-in color names that you can use. Run the command colors() to see the different options. An example use would be: col=c("grey5", "skyblue3")

By grayscale

You can specify a degree of grayness by using the gray(...) function and specifying numbers between 0 (black) to 1 (white). An example use for 4 colors would be: col=grey(0, 0.2, 0.5, 0.8)

By palette

The colorBrewer library contains custom palettes designed for various uses. You can see all the different palettes and their names by using the command display.brewer.all(). They fall into three categories:

• sequential palettes, starting from a fairly white color and getting progressively darker. Good for ordered categorical data.
• diverging palettes, with a middle neutral color and diverging in two directions. Good for ordered categorical data with a middle neutral value and negative/positive sides.
• qualitative palettes, with easy to distinguish colors, useful for nominal (unordered) categorical variables.

In order to use a palette, you would use the command brewer.pal(<n>, <name>) where <n> is the number of colors to choose, and <name> is the palette’s name, in quotes. For instance to use four colors from the “Accent” palette we would do: col=brewer.pal(4, "Accent")

histogram

histogram is used to produce histograms. The typical syntax would use a formula ~<variable> as the first argument, followed by a data=<dataset> option to specify the dataset to use.

You can also create multiple panels based on the values of a categorical variable, with a formula ~<variable>|<factorVariable>.

histogram(~black, data=counties, breaks=40, col="green")
histogram(~physhealth|genhealth, data=brfss)

barchart

barchart is used to produce bar charts. It is typically used by first doing a tally and then piping the result into barchart.

tally(~genhealth, data=brfss) %>% barchart(horizontal=FALSE)

tally(~genhealth|exerciseany, data=brfss, format="percent") %>% t() %>%
    barchart(auto.key=list(space="right"))