Introduction to R

6/26/15, 12:48 AMIntroduction to R
Page 1 of 80ﬁle:///Users/bencharoenwong/Dropbox/Citadel/R_overview/introToR-bencharoenwong.html#1
Introduction to R
Ben Charoenwong
June 25, 2015
R Markdown
This is an R Markdown presentation. Markdown is a simple
formatting syntax for authoring HTML, PDF, and MS Word
documents. For more details on using R Markdown see
http://rmarkdown.rstudio.com.
I wrote this document through R Studio, which is an editing
environment for R. It can also compile PDFs, HTML documents,
This document contains live R code that is evaluated when this
document is "compiled" (or rather, "knitted").
2/80

Why R?
R is a freely available language and enviroment for statistical
computing and graphics. It provides a wide variety of statistical
and graphical techniques.
Most statisticians are using R (though not all finance firms do). A
large number of add-on packages can be installed and run on R.
You rarely need to code anything from scratch yourself
R's syntax is very simple (so simple it's a bit like sloppy
programming). Since R is an interpreted language, not a compiled
one, it doesn't require building a complete program like in many
other languages (like C, Fortran, etc.)
·
·
·
·
3/80
How to install R
You can download R in different forms (Windows, Linux, Mac) at:
http://cran.r-project.org/
You can also get the most updated R version from there
For packages, to install, use install.packages() and to update,
use update.packages(ask = F).
·
·
·
4/80

Coding Environment
Once you initialize R, the programming console will appear.
The interface looks a little different in R Studio. Use whichever
you feel most comfortable with!
Code is written in scripts.
To create a new script, select "File" -> "New Script". The editor
looks standard.
You can set up other text editors to be able to interpret and run R
code too (won't get into this).
Code is saved in files ending with ".R", and code can be run using
the command source().
Don't forget to save!
·
·
·
·
·
·
·
5/80
Let's Start Coding

If/when you get stuck
Try the built-in help menu. For example, suppose I want to know
how the command "lm" works. The code below all have the same
effect:
·
?lm
help(lm)
help("lm")
Google for help. There is a large R user community on
StackOverflow and R-Blogger. Being able to find the most
efficient and readable code from these online resources is an
extremely valuable skill.
Email the TA (me) at: bcharoen@chicagobooth.edu, but only
AFTER you try steps 1 and 2.
·
·
7/80
Basic Commands
Let's jump into some basic commands.
Assign objects by using either "<-" or "=" (the former is more
robust)
·
a = 5
a
## [1] 5
-a
## [1] -5
8/80

Basic Commands
b = 4
b
## [1] 4
a + b
## [1] 9
9/80
Basic Commands
Example where "<-" is different from "=":
but {r, echo = T, eval = F} system.time(a = 5) returns an
error (you can try it).
system.time(a <- 5)
## user system elapsed
## 0 0 0
The "<-" always means assign, but "=" is also for passing in
arguments into a function. More on this later.
·
10/80

Basic Commands
R remembers everything that you assign (up to overwriting
variables) in memory. You can check what's in memory with
·
ls()
## [1] "a" "b"
11/80
Basic Commands
Just a tiny bit on style:
I use ";" to end a line in R. This is not necessary. Using ";" allows
me to sometimes compress multiple lines of code into one (we
can debate whether that is good or bad).
This is different from MATLAB! In MATLAB the ";" suppresses the
output.
I also usually use "=" to assign values, rather than "<-".
Mainly because I started programming in C++ and MATLAB.
·
·
·
·
12/80

Basic Commands
Just a tiny bit on style:
a = 5; a;
## [1] 5
b = 4; b + a;
## [1] 9
13/80
Basic Commands
Recall a = 5; b = 4; :
a - b;
## [1] 1
a * b;
## [1] 20
a / b;
## [1] 1.25
14/80

Basic Commands
Recall: a = 5; b = 4;:
a ^ b;
## [1] 625
exp(b);
## [1] 54.59815
log(b); # Default base "e". You can specify "base = 2", etc.
## [1] 1.386294
15/80
Basic Commands
Recall: a = 5; b = 4;:
min(a, b);
## [1] 4
max(a, b);
## [1] 5
abs(-a);
## [1] 5
16/80

Basic Commands
Recall: a = 5; b = 4;:
R doesn't like this kind of imaginary numbers. Need to be assigned
slightly differently
sqrt(a);
## [1] 2.236068
sqrt(-a);
## Warning in sqrt(-a): NaNs produced
## [1] NaN
17/80
Basic Commands
R also handles real and imaginary numbers. (You may never need to
use these though.)
x = 0 + 5i; temp = complex(imaginary = sqrt(5));
Re(x); Re(temp);
## [1] 0
## [1] 0
Im(x); Im(temp);
## [1] 5
18/80

Vectors & Matrices
Basic Commands: Vectors
Let's look at some vectors, and a case where R may work when
we don't want it to!
By default, operations in R are element-wise
We will look at matrix operations in a couple of slides
·
·
·
x = c(1:4); x; # by default: row vector
## [1] 1 2 3 4
y = c(1:3);
x * y
## Warning in x * y: longer object length is not a multiple of shorter object
## length
20/80

The rep command repeats the x-variable times-number of times
z1 = rep(5, 2); z1;
## [1] 5 5
z2 = rep(x = 5, times = 2); z2;
## [1] 5 5
z3 = rep(times = 2, x = 5); z3;
## [1] 5 5
21/80
The rep command repeats the x-variable times-number of times
z1 = rep(x = c(1,2,3), times = 2); z1;
## [1] 1 2 3 1 2 3
22/80

Basic Commands: Matrices
To create a matrix (note: Matrix != Array/List != Data Frame)
M = matrix(c(1:4),nrow = 2, ncol = 2);
M;
## [,1] [,2]
## [1,] 1 3
## [2,] 2 4
N = matrix(c(1:4),nrow = 2, ncol = 2, byrow = T);
N;
## [,1] [,2]
## [1,] 1 2
## [2,] 3 4 23/80
Transposing a matrix
M;
## [,1] [,2]
## [1,] 1 3
## [2,] 2 4
t(M);
## [,1] [,2]
## [1,] 1 2
## [2,] 3 4
any(t(M) != N); # element-wise checking whether M[i,j] != N[i,j]
24/80

Let's look at some matrix operations
M * N;
## [,1] [,2]
## [1,] 1 6
## [2,] 6 16
M %*% N;
## [,1] [,2]
## [1,] 10 14
## [2,] 14 20
25/80
What about subscripting matrices?
The command M[1,5] gives an error
Error in M[1, 5] : subscript out of bounds
M;
## [,1] [,2]
## [1,] 1 3
## [2,] 2 4
M[1,2];
## [1] 3
26/80

What about subscripting matrices?
M[1];
## [1] 1
M[4]; # bad style!
## [1] 4
M[6]; # note: no error! Unlike the subscript out of bounds
## [1] NA
27/80
Picking out rows and columns (used frequently)
M[1,];
## [1] 1 3
M[,1];
## [1] 1 2
28/80

We can also specify which rows we DON'T want
M;
## [,1] [,2]
## [1,] 1 3
## [2,] 2 4
M[-1,];
## [1] 2 4
M[,-1];
29/80
You can name rows and columns in matrices (same with
data.frame), so you can refer to names instead of indices
row.names(M); colnames(M);
## NULL
## NULL
row.names(M) = c("dave","pete");
colnames(M) = c("weight","height"); #note: function name inconsistent!
M
## weight height
## dave 1 3
30/80

You can name rows and columns in matrices (same with
data.frame), so you can refer to names instead of indices
M;
## weight height
## dave 1 3
## pete 2 4
M["dave","height"]; M[c("pete","dave"), "height"];
## [1] 3
## pete dave
## 4 3
31/80
Inverting some matrices may cause problems (numerically singular,
etc.)
M;
## weight height
## dave 1 3
## pete 2 4
solve(M);
## dave pete
## weight -2 1.5
## height 1 -0.5
32/80

To solve a system of equations of the form Mx = b
det(M);
## [1] -2
b = matrix(c(2,5) , nrow = nrow(M), ncol = 1); b;
## [,1]
## [1,] 2
## [2,] 5
solve(a = M, b = b); #solves for ax = b, a is square, b is column
33/80
Diagonal matrices
D = diag(c(1,2,3), nrow = 3) # much better than "eye" in MATLAB!!!
34/80

Diagonal matrices. Let's look at some potential pathological cases.
DANGER: THEY WORK!
diag(c(1,2,3), nrow = 3, ncol = 1)
## [,1]
## [1,] 1
## [2,] 0
## [3,] 0
diag(c(1:50), nrow = 2, ncol = 2);
## [,1] [,2]
## [1,] 1 0
## [2,] 0 2 35/80
Data Types

Basic Commands: Data Types
We use the command class() to verify the variable type.
class(M)
## [1] "matrix"
class(4)
## [1] "numeric"
class(c(1,2,3))
## [1] "numeric"
37/80
vec = c(1:5);
vec > 2;
## [1] FALSE FALSE TRUE TRUE TRUE
as.numeric(vec > 2);
## [1] 0 0 1 1 1
38/80

Like in many other languages, logicals and numerics are brethren.
as.logical(c(0,2,1,0))
## [1] FALSE TRUE TRUE FALSE
39/80
Let's look at some potentially pathological cases.
is.na(NA); is.nan(NaN);
## [1] TRUE
## [1] TRUE
class(NA); class(NaN);
## [1] "logical"
## [1] "numeric"
40/80

Some functions/operations only work on specific variable types
(cue: overloading concepts).
x = complex(real = 4);
class(x);
## [1] "complex"
as.numeric(x);
## [1] 4
class(as.numeric(x));
41/80
Some functions/operations only work on specific variable types
(cue: overloading concepts).
y = complex(imaginary = 5);
as.numeric(y);
## Warning: imaginary parts discarded in coercion
## [1] 0
42/80

Again, let's consider some potentially pathological cases.
class(as.numeric(NA));
## [1] "numeric"
as.logical(NaN);
## [1] NA
class(as.logical(NaN));
## [1] "logical"
43/80
"Combinding"

Basic Commands: "Combinding" Items
You can combine things of the same data type.
Command c() works for vectors
·
·
x = c(1,2,4);
c(x, 5);
## [1] 1 2 4 5
y = c(T, TRUE, F, FALSE);
c(y, NA, FALSE);
## [1] TRUE TRUE FALSE FALSE NA FALSE
45/80
Recall that logical and numeric are brethren, and that numeric is
more general. It will automatically be converted.
Command c() works for vectors·
c(F,10,2,T,F);
## [1] 0 10 2 1 0
c("hi","world","!");
## [1] "hi" "world" "!"
46/80

We can use cbind() and rbind() to combine
matrices/data.frames
·
M;
## weight height
## dave 1 3
## pete 2 4
rbind(M, c(1,2)); # a bit sloppy
## weight height
## dave 1 3
## pete 2 4
## 1 2 47/80
Pathological cases
·
·
rbind(M, 5);
## weight height
## dave 1 3
## pete 2 4
## 5 5
48/80

R will NOT NECESSARILY throw an error when using rbind() if
number of columns don't match.
Pathological cases·
rbind(M, c(1,2,3));
## Warning in rbind(M, c(1, 2, 3)): number of columns of result is not a
## multiple of vector length (arg 2)
## weight height
## dave 1 3
## pete 2 4
## 1 2
49/80
·
newRow = c(1,2);
rbind(M, newRow); # still a bit sloppy. also notice row.name
## weight height
## dave 1 3
## pete 2 4
## newRow 1 2
50/80

·
newRow = c(1,2);
rbind(M, newRow); # still a bit sloppy. also notice row.name
## weight height
## dave 1 3
## pete 2 4
## newRow 1 2
51/80
·
newRow = matrix(c(1,2), ncol = 2, nrow = 1);
rbind(M, newRow) # slightly better style
## weight height
## dave 1 3
## pete 2 4
## 1 2
52/80

·
# another difference between assign vs "="
rbind(M, newRow <- matrix(c(1,2), ncol = 2, nrow = 1));
## weight height
## dave 1 3
## pete 2 4
## 1 2
53/80
I include those I have used a lot in financial applications.
Other Useful Commands

Basic Commands: Additional
x = c(2,3,5,1,8);
length(x);
## [1] 5
sort(x);
## [1] 1 2 3 5 8
sort(x, decreasing = T);
## [1] 8 5 3 2 1
55/80
x;
## [1] 2 3 5 1 8
sum(x);
## [1] 19
cumsum(x);
## [1] 2 5 10 11 19
56/80

x;
## [1] 2 3 5 1 8
prod(x);
## [1] 240
cumprod(x);
## [1] 2 6 30 30 240
57/80
x;
## [1] 2 3 5 1 8
diff(x);
## [1] 1 2 -4 7
diff(x, 2);
## [1] 3 -2 3
58/80

x;
## [1] 2 3 5 1 8
rank(x);
## [1] 2 3 4 1 5
cut(x, breaks = 2);
## [1] (0.993,4.5] (0.993,4.5] (4.5,8.01] (0.993,4.5] (4.5,8.01]
## Levels: (0.993,4.5] (4.5,8.01]
59/80
Basic Application: Ranking Analysts
Here is an example:
data = data.frame(analystID = c(10001,10002,10003),
ret = c(0.05, -0.21, 0.62));
data;
## analystID ret
## 1 10001 0.05
## 2 10002 -0.21
## 3 10003 0.62
60/80

Basic Application: Ranking Analysts
Here is an example:
data$rank = rank(-data$ret);
data;
## analystID ret rank
## 1 10001 0.05 2
## 2 10002 -0.21 3
## 3 10003 0.62 1
61/80
This is probably what you will doing most.
Data Analysis with R

Set Working Directory
Should be pretty self-explanatory. It sets the default for
input/output interactions with R.
For Windows, the file path is with double backslashes :(
setwd("C:Usersbencharoenwong")
getwd(); #this gets the current working directory
## [1] "/Users/bencharoenwong/Dropbox/Citadel/R_overview"
setwd("/Users/bencharoenwong/"); #for Linux/Mac
63/80
Reading in Data
The most commonly used commands will be:
read.csv()
read.table()
R can read many other formats:
Excel using read.xls in gdata or xlsx
SAS using spss.get in Hmisc
SPSS using read.sas7bdat in sas7bdat
Stata using read.dta in foreign
SQL using RMySQL, RPostgreSQL
etc.
·
·
·
·
·
·
·
·
· 64/80

Generating Data
data = data.frame(GS = rnorm(100, mean = 0.2, sd = 0.4),
MS = rnorm(100, mean = 0.1, sd = 0.8));
head(data, n = 4); #default is n = 6
## GS MS
## 1 0.22541579 0.7166003
## 2 0.14879169 -0.1149048
## 3 0.07723751 0.5002469
## 4 -0.54295420 0.5789232
tail(data, n = 4);
## GS MS
## 97 -0.02045411 0.4689193
## 98 0.91976861 -0.9991719 65/80
Generating Data
You can draw random numbers from many distributions. They all
usually take form of r + <distribution>
You can include more distributions from other packages too.
runif (default U[0,1])
rnorm (default N(0,1))
rt (must set degree of freedom df)
rchisq (must set degree of freedom df)
·
·
·
·
66/80

Plotting
Let's look at plotting in the base package. (I don't really use this
anymore. I use the package ggplot2).
plot(x = cars$speed, y = cars$dist); # default scatter, ugly hollow point
67/80
Plotting
plot(x = cars$speed, y = cars$dist, main = "Dist vs. Speed",
type = "p", pch = 19); # look up "pch" online...it's a pain!
68/80

Plotting
type = "l", col = "red", xlab = "Speed", ylab= "Distance (km)");
abline(a = 0.5, b = 10, col = 5); # of the form ax + b
69/80
Plotting
type = "l", col = "red", xlab = "Speed", ylab= "Distance (km)");
abline(h = 20, col = "blue");
abline(v = 17, col = "green");
70/80

Statistics
The standard statistical functions work:
mean() computes sample mean
var(), or sd() which is just the sqrt(var())
median() (might not be in data. If coinciding with 2 numbers, will
take average.)
cov() computes sample covariance
cor() computes sample correlation, can specify pair-wise and
correlation type.
hist() makes histograms
summary()
No built-in function for the statistical mode!
·
·
·
·
·
·
·
· 71/80
Statistics
summary(cars);
## speed dist
## Min. : 4.0 Min. : 2.00
## 1st Qu.:12.0 1st Qu.: 26.00
## Median :15.0 Median : 36.00
## Mean :15.4 Mean : 42.98
## 3rd Qu.:19.0 3rd Qu.: 56.00
## Max. :25.0 Max. :120.00
mean(cars$speed); var(cars$speed); sqrt(var(cars$speed)) == sd(cars$speed);
## [1] 15.4
## [1] 27.95918
72/80

Statistics
hist(cars$speed, breaks = 10);
73/80
Linear Models
Basic regression models can be estimated with the command lm.
There are more general and special versions of this.
summary(lm(speed ~ dist, data = cars));
##
## Call:
## lm(formula = speed ~ dist, data = cars)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.5293 -2.1550 0.3615 2.4377 6.4179
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.28391 0.87438 9.474 1.44e-12 ***
74/80

This will go through the basic loops. This will NOT cover
the apply class of functions, nor advanced function
application tools like dplyr or data.table.
Programming with R
For Loops
Pretty self explanatory.
for (i in c(1:5)) {
print('hi');
};
## [1] "hi"
## [1] "hi"
## [1] "hi"
## [1] "hi"
## [1] "hi"
76/80

While Loops
Pretty self explanatory.
i = 0;
while (i < 5) {
print('hi');
i = i + 1;
};
## [1] "hi"
## [1] "hi"
## [1] "hi"
## [1] "hi"
## [1] "hi"
77/80
Conditionals
If, and if-else commands.
x = 5; y = 3;
if (x > y) {print("x more than y");}
## [1] "x more than y"
if (x <= y) {print ("x less than or equal to y");}
78/80

Conditionals
If, and if-else commands.
x = 5; y = 3;
if (x > y) {
print("x more than y");
} else {
print ("x less than or equal to y");
}
## [1] "x more than y"
79/80
Conclusion
R is a powerful tool!
Use R!
Depending on how much R you anticipate to use in the future,
decide whether to pay the fixed cost to learn new
packages/commands.
There are also a lot of shady packages out there…
Time to quit with q().
·
·
·
·
·
80/80

Introduction to R

Recommended

Recommended

More Related Content

Viewers also liked

Viewers also liked (15)

Similar to Introduction to R

Similar to Introduction to R (20)

Recently uploaded

Recently uploaded (20)

Introduction to R