Econometrics 3b

Computer Lab 1: `R` Intro

Resources (all available for free)

“An Introduction to R” ebook that starts from scratch and includes videos (for beginner R users) link
introduction by the developers of R (for R beginners with other programming experience) link
“R for Data Science” ebook for a hands-on introduction to data science in R with a particular focus on packages from the tidyverse (for beginner and intermediate R users) link
“Efficient R Programming” ebook on how to efficiently write code, and how to write efficient code (for intermediate and advanced R users) link
“R Packages” ebook on how to write and publish R packages (for intermediate and advanced R users) link
advanced tips and tricks with a nerdy presentation link
“Advanced R” ebook that goes deep, yet is easy to follow link

What is `R`?

open-source programming language developed for statistical analysis, computation, and visualization
R compared to Stata, SPSS, or SAS
- has no point-and-click interface
- has a steeper learning curve
- is more difficult to use if you “just want to run regressions”
- is much more flexible (useful when cleaning data, or when using more involved estimation procedures)
- has a vastly superior online support community, and learning resources (any question you can have is probably answered on stackoverflow)
- is a “fully fledged” programming language, i.e. can do more
R compared to Matlab, Octave, or Julia
- is more focussed on statistical analysis (i.e. it is easier to do)
- has better data-handling capabilities
- is typically slower
R compared to Python
- is more focussed on statistical analysis (i.e. it is easier to do)
- has better out-of-the-box data-handling capabilities
- is less of an all-round language

`R` and `RStudio`

R is the language in the background, RStudio is “optional”
RStudio is an IDE (integrated development environment), and a very good one at that (de facto standard)
use it to write R scripts, (interactively) execute code, look at plots, and look at data
note that correct R code can run without RStudio

Download R here

Download RStudio here

Using `R`

`RStudio`

default window arrangement
- left is console, the current R instance
- top right is environment (current variables), history (previously executed commands)
- bottom right is very important, help, plots, and more
highly recommended working with projects
- forces you to keep code and data structured
- idea is to keep everything related to a “project” in a single folder and subfolders
  - both scripts and data
  - allows you to use relative file paths
  - reproducible by just sharing that folder

Overpowered calculator

1 + 1

## [1] 2

2 - 1

## [1] 1

2 * 5

## [1] 10

4.4 / 2

## [1] 2.2

2^5

## [1] 32

log(10)

## [1] 2.302585

exp(2)

## [1] 7.389056

Assignments

x <- 2
x

## [1] 2

3 * x

## [1] 6

log(x)

## [1] 0.6931472

assignment possible using <- or =
- R convention is to use <-
- = also used for keyword arguments in function calls (more later)

Data Modes

three data modes: numeric, character, and logical

y <- "tjena"
z <- TRUE

mode determines what can be done with the object

try(log(y))

## Error in log(y) : non-numeric argument to mathematical function

Data Structures

four important data structures: vector, matrix, data.frame, and list
- special cases:
  - a scalar is a length-1 vector
  - a matrix is a 2-dimensional array
  - a data.frame is a special list

`vector` and `matrix`

can contain only elements of one mode

vec1 <- c(1, 2)
vec2 <- c("tja", "tjena")
vec3 <- c(TRUE, FALSE)

matr1 <- matrix(c(1, 2, 3, 4), nrow = 2)

`list`

can contain anything (even more lists)

list1 <- list(2, "tja", TRUE, list("bla bla"))

list is very versatile and flexible, but you have to be careful
more or less irrelevant for us now

`data.frame`

workhorse for data analysis
like a matrix, but data modes can differ between columns

df1 <- data.frame(x = c(1, 2), y = c("tja", "tjena"))
df2 <- data.frame(x = vec1, y = vec2)

df1

##   x     y
## 1 1   tja
## 2 2 tjena

df2

##   x     y
## 1 1   tja
## 2 2 tjena

Subsetting

access individual elements, or ranges of elements using []
R is one-indexed: first element of vector is x[1] (unlike e.g. Python, where it is x[0])

# shorthand to create a vector containing 1, 2, ..., 10
x <- 1:10

x[3]

## [1] 3

x[5:8]

## [1] 5 6 7 8

matrix subsetting similar, use [row, column]

y <- matrix(1:4, nrow = 2, byrow = TRUE)

y[1, 1]

## [1] 1

y[, 2]

## [1] 2 4

note that the last command returns a “directionless” vector; we extract a column of the matrix, but get a “standard” vector (i.e. not row or column)

Time Series

R comes with built-in time series capabilities

x <- ts(1:12)
class(x)

## [1] "ts"

# above is kind of useless, could have just as well used x <- 1:12

# useful feature is to add date and frequency info
x <- ts(1:12, start = c(2022, 1), freq = 12)

# note x-axis label
plot(x)

using ts() not strictly necessary, but useful for understanding data and easier plotting

Functions

every operation is a function
performs an action, or sequence of actions
has a name, and arguments
takes objects as inputs
e.g. c() takes two scalars as inputs, and combines/concatenates them to one vector
all functions have help-pages, describing what the function does, input arguments, and output

help(c)
?c

examples

x <- seq(1, 5, by = 1)
sum(x)

## [1] 15

mean(x)

## [1] 3

r <- rnorm(10)

functions have default values, look at help page of rnorm
functions are organized in packages

Packages

look at base package in RStudio
great advantage of open-source languages like R or Python is huge universe of user written packages
anyone can write and publish a package
CRAN (Comprehensive R Archive Network) is the central repository

Is a package “good”?

when googling for how to do things in R, sometimes you find very particular packages that promise to help you
- but is a package “good” (i.e. bug-free and does what it says)?
- google “cran [package name]”, and click the first link
- e.g. “cran dynlm” points us to here
- check the packages metadata:
  - version 0.3-6: be alert, early version of the package
  - published 2019: OK
  - author Achim Zeilis with email address at @r-project.org: good sign
- another example: “cran did” here
  - advanced version number
  - published very recently
  - authored by one of the developers of the estimator himself -> good sign
- be aware of low version numbers, old publication dates, and random authors, the code might contain bugs
- CRAN runs technical checks of the code before publication (i.e. do the functions run without errors/warnings), does not check if the numbers are correct

Console and Scripts

essentially two modes of operation: interactive with the console, or “organized” with scripts
R scripts are text documents (like .txt, .csv, .py, .do, …) with file extension .R
write your code sequentially in the script, and execute either all at once from the command line, or line-by-line in RStudio

Hands-On

Installing Michael’s `TS` package

open RStudio, create project (new directory, call it “Lab-1” or something)
- remember the location
download .tar.gz file from athena, put it into the folder just created for the project
go to “Console” in RStudio, execute list.files()
- should show "TS_1.0-2022.1.tar.gz"
execute

install.packages("urca")
install.packages("vars")
install.packages("TS_1.0-2022.1.tar.gz", type = "source", repo = NULL)

Writing our first script

create new R script, save it in the project directory
add comment at top of file describing what this is
load TS package
load KPIF data
calculate average inflation

# Calculate average inflation rate in Sweden

library(TS)

data("KPIF")

print("The average inflation rate in Sweden is")

mean(KPIF)

save file
in RStudio go to bottom left, “Terminal”
type RScript "[filename].R", enter
the script is then run in batch mode
- Michael requires this for all assignments and fails you if it does not work, so test this!
- reasons for it not to work (on Michael’s computer):
  - you use setwd() with absolute paths in the script (specify all paths relative to project directory)
  - the order of commands in wrong (need to load the package first, before we can load the data)
  - a simple coding error

More?

RStudio with integrated tutorials through learnr package
- I have not used it, can not judge
- introduces the tidyverse, which is a particular “style” of using R
- access in RStudio in top-right
swirl package
- interactive in the R console
- that’s how I learned the basics

install.packages("swirl")
library(swirl)
swirl()

Econometrics 3b

Thore Petersen¹

24.03.2022

Computer Lab 1: `R` Intro

Resources (all available for free)

What is `R`?

`R` and `RStudio`

Using `R`

`RStudio`

Overpowered calculator

Assignments

Data Modes

Data Structures

`vector` and `matrix`

`list`

`data.frame`

Subsetting

Time Series

Functions

Packages

Is a package “good”?

Console and Scripts

Hands-On

Installing Michael’s `TS` package

Writing our first script

More?

Econometrics 3b

Thore Petersen1

24.03.2022

Computer Lab 1: R Intro

Resources (all available for free)

What is R?

R and RStudio

Using R

RStudio

Overpowered calculator

Assignments

Data Modes

Data Structures

vector and matrix

list

data.frame

Subsetting

Time Series

Functions

Packages

Is a package “good”?

Console and Scripts

Hands-On

Installing Michael’s TS package

Writing our first script

More?

Thore Petersen¹

Computer Lab 1: `R` Intro

What is `R`?

`R` and `RStudio`

Using `R`

`RStudio`

`vector` and `matrix`

`list`

`data.frame`

Installing Michael’s `TS` package