PRELIMINARIES
A Warning.
Take my word for this. Not everything will make sense to you at first! If there is anything that puzzles you, don't worry. You will probably see it illustrated in the tutorials to come.
What is R?
First, there was S. The S statistical programming language was developed in the late 1970's, primarily by John Chambers at Bell Labs. S was first distributed outside of Bell Labs in 1980, and by 1988 the "New S Language" had become available. This was the basis for the commercial version called S-Plus. By 1990, S or S-Plus was in widespread use by statisticians.
In the early 1990's, Ross Ihaka and Robert Gentleman, of the University of Auckland in New Zealand, wrote a teaching version of S/S-Plus and named it R. In 1995 the source code for R was released as open source under the Gnu Public License. People who are unfamiliar with this concept are urged to click those links and read all about it. Since that time, R has become one of the most powerful and versatile statistical software programs available at any price.
R is a statistical computing environment. Primarily, it is a programming language, but one containing a very large number of statistical functions. These functions can be used to perform complex statistical analyses interactively, or they can be included in larger scripts and programs to accomplish even more complex tasks. R also has very elaborate graphical capabilities, allowing the production of publication-quality graphics.
R is free software. To get R, simply go to the R Project homepage and download it. Click on the CRAN link on the lefthand side of the page under downloads. (CRAN stands for Comprehensive R Archive Network.) R is available for Windows (Windows 95 and later), Mac OS X, and Linux. As of this writing (26 July 2010), the latest version is 2.11.1.
Once you have R, you may do almost anything with it you please. You can install it on as many computers as you want. You can examine and modify the source code. You can even resell it, as long as you make the source code available. Unlike commerical software such as SPSS, you are not paying an exorbitant fee to "rent" a restricted-use copy. R is free!
The R Prompt.
Start R like you would start any other program on your computer. In Windows, the R installer will place a shortcut icon on your desktop. On the Mac, you will have to drag an icon from your Applications folder to the dock. In Linux, R runs from a terminal window. You can also run R from either terminal.app or an X11 terminal on a Mac. I wouldn't recommend this in Windows, however.
R is command-line driven. This means you type your commands at a prompt rather than hunting and clicking through menus. This is much faster and more versatile than a GUI (graphical user interface), but it does take some getting used to for those of you who rarely take your hands off your mouse.
R GUIs are in development. Google up "R Commander" for example. The S-Plus commercial version also has a GUI, but be ready to pay dearly for it, in cash money, that is. (Note: S-Plus is sold by www.insightful.com, which is now owned by TIBCO and appears to have been rebranded as Spotfire S+.)
The R prompt is a greater than symbol, >. When you see it, start
typing. If your command is a long one and breaks onto the next line, or
if you hit Enter without completing a command, R will prompt you on the
next line with a plus sign (+). This means "you ain't done yet. Gimme
more." This often happens when you don't close a parenthesis. For
example...
> oneway.test(weight ~ group, data = PlantGrowth # Note: press Enter here
+ # Type ) and press Enter here
You neglected to close the open left parenthesis. Close it and hit Enter and
everything will be fine. R is not too picky about how you type your commands as
far as spacing is concerned, but it will insist that parentheses (and quotes)
come in left and right pairs. By the way, go ahead and type the commands that
you see here into R. The data objects required to make these examples work are
built in to R for educational purposes. Just remember that R is case sensitive,
so in addition to parentheses, you will also have to get capitalization (and
spelling) correct.
Note: In the newer versions of R on the Mac, parentheses (and quotes) are closed for you. I.e., as soon as you type the left one, the right one also appears. This can be very convenient; in some cases, it can also be annoying. A common trick that programmers use is to type left and right parentheses at the same time, then backspace and type whatever goes between them.
Further note: In some versions of R, especially those on Windows, when your typing reaches the right side of the window, R does not break the line but scrolls the window to the right. Once again, convenient and annoying. You can insert a break yourself by hitting the Enter key, if you prefer the window not to jump around.
Quitting R.
Best to know this right away, I suppose. To quit R, type...
> quit()
...and then press the Enter (or Return) key. You always tell R to execute a
command line by pressing the Enter key. By the way, q( ) is a shortcut for "quit" and will also work.
The parentheses are mandatory. R will ask if you wish to save your workspace. Say yes if you want to save any of the objects you created while working in R. (See the next tutorial for "objects.")
Case Sensitive.
R is case sensitive. "My_data" and "my_data" are not the same objects. Nor are Anova( ) and anova( ) the same functions (commands). The most common reason I get error messages is capitalizing where I shouldn't have, or not where I should have. If you get an "unknown object" or "function not found" error, check your capitalization first. And then check your spelling!
This might also be a good place to point out that R does not like spaces in the names of things. Use a dot or an underline character instead. Thus, "my.data" and "my_data" are fine (and different) names for a data set, but "my data" is not allowed. "MyData" and "myData" are also allowed (and different) names for data objects. But don't use a dash: "my-data". R will think you mean subtraction: "my" minus "data". You should follow these rules not only for named data objects you create while using R, but also in files you save with R. I.e., don't put spaces in filenames. R will work around it if you do, but it's just better to avoid the hassle to begin with.
Comment Lines.
R code, like any good programming code, can be commented. Any line or
partial line in R beginning with the hash or pound symbol (#) is a
comment, or note, and R will ignore it. You can use these to annotate
your analysis or make notes to yourself (or others).
> # This is just a note.
> ### And so is this.
> summary(rivers) # And so is this.
> # In the last command line, R will summarize the "rivers" variable and
> # ignore all these notes. Try it and see!
If you are going to save your R session (explained in a future tutorial),
liberal comments will help you remember what you've done.
R Functions.
Almost every command you issue in R will take the form of a function.
Functions have the following syntax...
> function.name(arguments, options)
For example...
> mean(islands, na.rm = T)
This function takes the mean of a variable called "islands", the argument. The
option, na.rm=T tells R to remove missing values,
if any, before doing so. R will not calculate the mean of a variable that has
missing values unless you tell it that the missing values should be removed.
Unlike some statistical software, R is clever enough to know that variables
with missing values don't have means.
The parentheses are manditory, even if there are no arguments or
options given. For example...
> ls()
If you type the name of a function without the parentheses, R will show you how
the function is coded; i.e., it will print out the programming code for the
function. I mention this because it can be frightening to the unwary! It's
harmless though. You haven't broken anything. Try it! Type ls and leave off the parentheses. By the way, you can
scroll the R console window with your mouse to see the stuff that scrolled off
the screen. Don't expect it to make much sense to you though.
A Taste--Before You Get Too Impatient!
If you haven't already, fire up R and get to typing.
For educational purposes, R has a large number of data sets built into it. I will use these to illustrate a few things R can do. When you start R, you will get a screen something like this...
R version 2.10.1 (2009-12-14) Copyright (C) 2009 The R Foundation for Statistical Computing ISBN 3-900051-07-0 R is free software and comes with ABSOLUTELY NO WARRANTY. You are welcome to redistribute it under certain conditions. Type 'license()' or 'licence()' for distribution details. Natural language support but running in an English locale R is a collaborative project with many contributors. Type 'contributors()' for more information and 'citation()' on how to cite R or R packages in publications. Type 'demo()' for some demos, 'help()' for on-line help, or 'help.start()' for an HTML browser interface to help. Type 'q()' to quit R. [Previously saved workspace restored] >
The R core team and R developers the world over are volunteers doing
this, well, maybe not entirely out of the goodness of their hearts, but
without much compensation. They would like you to cite them if you use
R for data analysis. To see how to do this, issue the following command...
> citation()
Congratulations! You're an R user! (And a special note to my students using
this tutorial: You are REQUIRED to cite R if you use it for your data analysis
or graphics.)
To give you a bit of an idea of what R is capable of graphically, try this...
> demo("graphics")
You will get a message telling you to "Type <Return> to start". This means
press the Enter (or Return) key. When you do, a gray graphics window will open
and some stuff will print out in the R Console window. You can ignore how these
things are coded (for now). Just watch the graphics window. In the R Console it
says "Waiting to confirm page change..." This means press Enter again. If
nothing happens, be sure the graphics window has focus (i.e., click in the
graphics window somewhere), and then press Enter again. Each time you press
Enter, a new graph will appear. There are about 11 in all. R's graphics
capabilities are pretty impressive, and this is just a small sample. When you
are done admiring the graphs (when the graphs stop changing upon hitting Enter),
close the graphics window like you would any other window on your system.
This little exercise has placed a lot of stuff in your workspace (to be
explained later). We should clean that up. WARNING: The command I am about to
tell you to execute will erase everything in your workspace, and that will be
permanent. If you've been using R already, and there is something in your
workspace you want to save, save it now!
> ls() # List the contents of the workspace.
> rm(list=ls()) # This completely clears the workspace.
> ls()
character(0) # This means "nothing to see here"
In Windows and on the Mac, you can also manage your workspace via menus. In
Windows, use your mouse to put down the "Misc" menu and choose "List objects"
to see the contents of your workspace. (It seems to me that typing ls( ) is a lot easier!) Pull down "Misc" and choose
"Remove all objects" to clear the workspace. On the Mac, you use the
"Workspace" menu to accomplish the same things. Choose "Show workspace" and
"Clear workspace" to do so. Now let's look at some data.
As a brief appetite whetting, let's look at a data set called
"HairEyeColor". This is a table showing a crosstabulation of hair color, eye
color, and sex (gender) for 592 statistics students. Type this, and don't type
the command prompt--R has already supplied that for you. Remember to press
Enter to execute the command line...
> data(HairEyeColor) # Put the data in your workspace. Not actually necessary!
> HairEyeColor # Remember to watch your capitalization! And spelling!
Since "HairEyeColor" is already a defined object (more on this later),
typing its name will cause R to display it. This is actually a shortcut for the function...
> print(HairEyeColor)
...which does the same thing. You should now be looking at a three-dimensional
(4x4x2) crosstabulation in your R console. The following command will show you
the names of the variables as well as the levels of them...
> dimnames(HairEyeColor)
A chi-square test for independence of all factors in the table (log linear
analsysis) can be done as follows...
> summary(HairEyeColor)
Number of cases in table: 592
Number of factors: 3
Test for independence of all factors:
Chisq = 164.92, df = 24, p-value = 5.321e-23
Chi-squared approximation may be incorrect
>
Because the p-value is so small, R has printed it in scientific notation:
5.321x10-23. You need to know this notation to use R. Whenever you
see a number followed immediately by a lower case "e" and then another number,
that "e" part means "times ten to this exponent". If the exponent is positive,
this is a very large number; if negative, a very small number (i.e., very
close to zero). If you are unfamiliar with or don't remember scientific
notation for large and small numbers, you will need to learn it. Click this
link for a brief lesson: Scientific
Notation.
The null hypothesis that all factors are independent (i.e., no interactions between any factors) is rejected. If you are getting error messages instead of statistical output, remember: R is case sensitive. HairEyeColor and haireyecolor and NOT the same thing. (Note: The warning message "Chi-squared approximation may be incorrect" means there are expected frequencies less than 5, for those of you who know what that means.)
You are not to worry about the details of this syntax at this point. This
is just for show! To collapse over one or more of the factors in this table,
you can do one of these...
> margin.table(HairEyeColor, c(1,2)) # collapses over sex (i.e., displays vars 1 & 2)
> margin.table(HairEyeColor, c(2,3)) # collapses over hair color
> margin.table(HairEyeColor, c(1,3)) # collapses over eye color
> margin.table(HairEyeColor, c(1)) # collapses over eye color and sex
Suppose we wished to perform an ordinary (i.e., two-way) Pearson
chi-square test of independence on hair color and eye color just for
the men. Here's how to do it (and once again, just for show so don't worry about
memorizing all this syntax, which probably won't make much sense to you at this
point)...
> chisq.test(HairEyeColor[,,1])
Pearson's Chi-squared test
data: HairEyeColor[, , 1]
X-squared = 41.2803, df = 9, p-value = 4.447e-06
Warning message:
In chisq.test(HairEyeColor[, , 1]) :
Chi-squared approximation may be incorrect
The test indicates a relationship between hair color and eye color. However,
there is a warning that the result may not be accurate. This warning occurs
when the expected frequencies may not be high enough to make the chi square
approximation accurate. In this case there is one cell of 16 with an expected
value less than 5, and it has EF=4, so I wouldn't worry too much about it. To
see these expected values, type...
> chisq.test(HairEyeColor[,,1])$expected
Eye
Hair Brown Blue Hazel Green
Black 19.67025 20.27240 9.433692 6.623656
Brown 50.22939 51.76703 24.089606 16.913978
Red 11.94265 12.30824 5.727599 4.021505
Blond 16.15771 16.65233 7.749104 5.440860
Warning message:
In chisq.test(HairEyeColor[, , 1]) :
Chi-squared approximation may be incorrect
To see the data table in a different format (called a flat table), try
this...
> ftable(HairEyeColor)
For yet another format, try this...
> as.data.frame.table(HairEyeColor)
Okay, that's enough whetting for now. There are a few more details we
need to cover before we can get down to data analysis.
Some Definitions.
workspace
The workspace is the area of your computer's memory (RAM) where R is
storing the variables and other things you are currently working with.
To see these objects...
> ls()
[1] "HairEyeColor"
The ls( ) function stands for "list the
objects in my workspace". An equivalent function is...
> objects()
Your output may be different if you've been fooling around with R or not
cleaning out your workspace after some of the above examples. If the output
says character(0), this means there is nothing in your workspace because you
haven't created or stored anything there. If you do this...
> x = 7
...an object called "x" is created in your workspace (and any older object
named "x" is overwritten, so be careful). We can see it...
> ls()
[1] "HairEyeColor" "x"
To remove an object from your workspace, use either rm( ) or remove( )...
> rm(x)
> ls()
[1] "HairEyeColor"
It's a good idea to keep your workspace clean. When you have finished using an
object, remove it so it doesn't get in the way when you begin your next
analysis. Be aware though that in R removing stuff from your workspace is
permanent. It cannot be recovered from the trash. A future tutorial will deal
with how to save a permanent copy of something.
There is one more consequence of your workspace being in RAM. If the power
goes out, or your laptop battery goes dead, your workspace is gone! If you're
worried about this, you can always do a quick save of your workspace as follows
...
> save.image()
This saves a copy of your workspace to the working directory in a file called
".RData".
working directory
This is a directory (or "folder" for those of you who have been spoiled
by Windows) on your hard drive or other storage device where R will
look when you ask it to read in a file. It is also where R will try to
save something if you ask to save a copy. To find out what's in your
working directory, type...
> dir()
[1] "Jordan_Data.xls" "BandK.csv"
[3] "R_functions" "Rspace_071025"
[5] "Rspace_test" "Sex2.csv"
[7] "Smith Logic file.xls" "UCB.saved"
[9] "age_chd.txt" "coronary.csv"
[11] "data_folder" "death_penalty"
[13] "elephants.txt" "elephants_analysis.odt"
[15] "elevCA_smoke_chd.txt" "eysenck.csv"
[17] "garafola_mullet" "hack_altered.csv"
[19] "hack_altered.sav" "hackett_data.csv"
[21] "logic_file.csv" "police_log.saved"
[23] "testdir"
>
First, take a moment to notice how R numbers things it prints to the console
(screen). In this output there are 23 objects (files), and each of them has a
number. Only the first one on each line of output is numbered, however. This
number appears in square brackets at the beginning of the line. So
"Jordan_Data.xls" is the first item in this output, "BandK.csv" is the second,
"R_functions" is the third, and so on. Remember this. It will be important
later.
You can identify your working directory as follows...
> getwd() # get pathname to my working directory
[1] "/Users/billking/Rspace"
It's a good idea, in my opinion, if you don't use your home directory (in
Windows XP it's My Documents, in Vista and I suspect Windows 7 it's your home
folder, on the Mac and in Linux it's your home folder) as the R working
directory. So, as you can see in the above example, I have created a directory
called "Rspace" inside the home directory (or inside My Documents in XP) to use
with R. I changed to it, and then viewed its contents, like this (don't type
these commands just yet)...
> setwd("Rspace")
> dir()
NOTE to everyone: In the tutorials that follow, I will assume you are
in a working directory called Rspace. So create that working directory
now. If you just started R, type the following commands. If you've been
fooling around in R for awhile, quit R and restart it, then type the
following commands...
> system("mkdir Rspace")
> setwd("Rspace")
> getwd()
[1] "/Users/billking/Rspace"
> ls()
[1] "HairEyeColor"
IMPORTANT: You only have to do the "system" command once per lifetime! It
creates the directory (folder) "Rspace", and once it's created it will always
be there (unless you delete it). DON'T do this command the next time you use R.
You can accomplish the same thing by using your operating system to create a
new folder inside your R working directory.
These commands should work in all three major OSes: Windows, OS X, and Linux. Notice when I switched to the new working directory, R brought my old workspace with me. That can be very convenient. It can also be a nuisance. Don't worry about it now. Just create the "Rspace" directory and switch to it.
FURTHER NOTE to Windows users: This "should" work in Windows, but it didn't when I just tried it. Windows is a pretty awkward OS when it comes to doing anything useful without a mouse. So if you got an error or warning message, do this. Quit R. In Windows, open your "My Documents" folder (in XP) or your home folder (in Vista or Windows 7). Right click on a blank space inside the folder. Choose "new folder" from the pop-up menu. Name the folder "Rspace". Restart R. Switch to the folder using setwd("Rspace"). You should be ready to go. Use getwd( ) to confirm that things went as intended.
From here on, at the beginning of every tutorial, you should begin by executing the command setwd("Rspace").
search path
When you ask R to do a calculation on an object, R
will look in the search path to find that object. It will also look in the
search path to find the function you are asking to execute. If it fails to find
either one, you'll get an error message like "object not found" or "function
not found". To see your search path...
> search()
[1] ".GlobalEnv" "package:stats" "package:graphics"
[4] "package:grDevices" "package:utils" "package:datasets"
[7] "package:methods" "Autoloads" "package:base"
>
This is the Windows result, but it will be similar in any other OS. The first
entry is your "global environment," also called your workspace. This is the
first place R will look for anything you ask it to use. The rest are various
tools and packages. We will find out how to add places to your search path in
future tutorials.
Lot's of stuff to remember, right? Don't worry too much about it right now. Working with something is the best way to learn it. If you use it, the knowledge will come! (Okay, apologies to W. P. Kinsella for that one!)
Getting Help.
To see a manual page in R for any function, type...
> help("function.name") # a shortcut is ?function.name
For example...
> help("mean")
I should point out that these help screens will open in separate windows on the
Mac and in Windows. These windows can be manipulated with the mouse just like
any other window on your screen. When you're done looking at, click the
appropriate button to close it. In Linux, the help screens appear in-line
with your R session in the R Console. To get back to your command prompt,
press q (lower case Q).
These manual pages are intended for experts and can seem inpenetrable until
you learn a little more about R. Don't worry about them for now. However, if
you're daring and want to see a worked example, try this...
> example("mean")
(In my opinion, many of the worked examples this function produces are
unnecessarily complex. They are definitely not intended for beginners, that's
for sure!)
If you don't know the name of the function you want, there is a way around
that. For example, suppose you want to calculate a median but don't know the
function to do so. Try this...
> help.search("median")
This will give you a list of functions, all of which have something to do with
the median. From looking through this list, you should be able to spot the
median( ) function. Most R commands are this
cryptic!
If you are looking for a function that does a "mean-like sort of thing",
and you're not quite sure what it's called, but you're pretty sure that "mean"
is part of its name, do this...
> apropos("mean")
The output will be every R function (in the search path) that has "mean" as
part of its name.
There are also R manuals online (and they also come with the download,
so you have them already on your hard drive). The two most important
ones are "An Introduction to R," and "R Data Import/Export." They can
be found here:
http://www.cran.r-project.org/manuals.html
By the way, CRAN stands for Comprehensive R Archive Network. There is
several tons of useful stuff online there.
A Final Preliminary Word or Two.
Don't expect to understand R all at once! This is a full-featured statistical programming language and analysis environment. You will never understand it all. R will do everything from 2+2 to factor analysis and generalized linear (and nonlinear) models. If you need it done, R will probably do it. I recently had to use a relatively new technique called generalized estimating equations. There aren't many software packages out there that will do it, but R will. I had to download an optional package from the CRAN site, but that's easy enough to do.
If there is something in these tutorials that puzzles you, make a note of it and move on. Ask someone when you have a chance, or wait for it to come up again in a later tutorial. Perhaps it will be explained more fully there. Or try the help (manual) page, but don't pin your hopes on those just yet. Reading those is a skill in itself. It took me quite awhile to get used to them.
revised 2010 July 26