AN INTRODUCTION TO R FOR MANAGEMENT SCHOLARS

Transcription

1 AN INTRODUCTION TO R FOR MANAGEMENT SCHOLARS 24 January 2017 Stefan Breet breet@rsm.nl

2 TODAY What is R? How to use R? The Basics How to use R? The Data Analysis Process

3 WHAT IS R? AN OVERVIEW

4 WHAT IS R? R is a language and environment for statistical computing and graphics R is an integrated suite of software facilities for data manipulation, calculation and graphical display. It includes (R-Project.org, 2017): an effective data handling and storage facility, a suite of operators for calculations on arrays, in particular matrices, a large, coherent, integrated collection of intermediate tools for data analysis, graphical facilities for data analysis and display either on-screen or on hardcopy, and a well-developed, simple and effective programming language which includes conditionals, loops, user-defined recursive functions and input and output facilities.

5 THE HISTORY OF R R in it s current form is a dialect of the programming language S First version of S is distributed outside of Bell labs S is developed by John Chambers and colleagues at Bell Labs as a statistical analysis environment for internal use Two books are published and the source code is licensed by AT&T for educational purposes

6 THE HISTORY OF R R in it s current form is a dialect of the programming language S Ross Iahak and Robert Gentleman create R at the University of Auckland and closely 1988 modeled it on S The system is rewritten in C and resembles the system we First public announcement of R use today. Features like functions are introduced and are described in the book The New S Language. Iahak & Gentleman make R free to use under the GNU General Public Licence

7 THE HISTORY OF R R in it s current form is a dialect of the programming language S The R Core Group is formed (including people associated with S-Plus), which controls R s source code R version is released R version is released

8 THE POPULARITY OF R R s popularity is rapidly increasing IEEE Spectrum s Programming Language Ranking C Java Python C++ R C# PhP JavaScript Ruby Go - Matlab SAS IEEE Spectrum Ranking Source: IEEE Spectrum Source: Muenchen, 2016

9 COMPANIES THAT USE R R is popular outside of academia as well

10 ADVANTAGES OF USING R Why do scientists, data analysts and companies like R? It is more than a piece of statistical software: it s a programming language! You can create your own objects, functions and packages. It s free and open source. Everybody can access the source code, build extra features (objects, functions, packages), detect bugs and solve them. It s extremely versatile. It is the most comprehensive statistical analysis environment available. Ranging from the most basic statistical tests to the most complex analyses or data visualisations, R can do it.

11 ADVANTAGES OF USING R Why do scientists, data analysts and companies like R? It makes Reproducible Research easy. You can easily save, edit and share the code behind your analyses so others can reproduce them. Version control (via Github or Subversion) is easy to implement. Training & Support. R has a huge community of users and contributors, which makes it easy to find support if you have question. There are plenty of online courses and resources available if you want to learn more. Data Visualisation. R has by far the best data visualisation tools. It s easy to create beautiful plots, images and figures even if you want to make them interactive or dynamic.

12 DISADVANTAGES OF USING R R has a couple of disadvantages compared to other statistical software R has a steep learning curve in the beginning Once you master the basics, however, learning the advance stuff is easy It can be relatively slow and more complex than other programming languages (such as Python) The data you can load into R is limited to the size of your computer s working memory (RAM) Rule of thumb: don t use datasets larger than half the size of your working memory (RAM) However, packages have been developed that can handle big datasets No quality control on every package

13 R VS. STATA Both are popular among academics R is a programming language, while STATA is a software package R is free and open source, while STATA has to be purchased (and is expensive) STATA serves as a reliable basis (quality of applications is secured by the company), while R s range of different applications is much larger R has more users and therefore a larger support community

14 R VS. PYTHON Both are among the most popular programming languages R is designed for academics and data analysts, while Python is designed for programmers (that might want to analyse data or apply statistical techniques) R focuses on better and user friendly data analysis, statistics and graphical models, while Python focuses on productivity and readability R is a great place to start: you can access Python via the rpython package You can find a comparison here

15 PACKAGES You can extend R by installing packages Packages are collections of R functions, data and compiled code Packages are stored in the library R comes with a standard set of packages and others are available for download Before you can use a package, you have to load it into the global environment You can load packages (i.e., sets of functions) with library( package.name )

16 EXAMPLES The people behind RStudio have published several well-developed packages RMarkdown Create dynamic reports Shiny Create interactive web applications ggplot2 Create beautiful plots

17 MORE EXAMPLES A package exists for every type of statistical analysis Analysis Regression Analysis (LM & GLM) Meta-Analysis Structural Equation Modeling Survival Analysis Psychological, Psychometric & Personality Research Text Mining / Analysis Mediation Analysis Moderation Analysis Network Analysis Package stats metafor, metasem lavaan, openmx, SEM survival psych tm mediation pequod igraph, sna

18 HOW TO USE R THE BASICS

19 ALWAYS START WITH RSTUDIO RStudio makes R easier to use The R Console The RStudio Interface

20 RSTUDIO It includes a code editor, debugging tools and visualization tools Code Editor = Edit and save your scripts Console = Quickly try out code + view the output of your code Workspace Workspace = Overview of all the R objects loaded into the working memory of your computer. Plots = Preview of your plots + overview of the files in your working directory + list of the packages on your computer + help documentation Code Editor Console Plots The RStudio Interface

21 TIDY DATA Adhere to the principles of Tidy Data to ensure high-quality code and analyses A dataset is said to be tidy if it satisfies the following conditions: 1. Each variable forms a column. 2. Each observation forms a row. 3. Each type of observational unit forms a table. Tidy data makes it easy to carry out data analysis. More information: Wickham, H Tidy Data. Journal of Statistical Software, 59(10).

22 TIDY DATA (CONTINUED) Make sure that others can understand your code Always make sure that the names of variables, values and functions are descriptive and human-readable Be consistent: use lower case and do not separate words by a space but use a dot (.) or underscore (_) instead For example: entrepreneurial_behavior instead of EntBehav or EB log_tenure instead of Tenure(Log) firm_performance instead of Firm Performance male / female instead of 0 / 1

23 GETTING STARTED Basic calculations & objects R is basically a calculator on steroids. Run the following commands in the RStudio console: 4+4 6*6 You can store a value as an object. You can create new objects with te assignment operator <- : x <- 4 y <- 6 8/ ^2 All R statements where you create objects have the same form: object_name <- value

24 OBJECTS & STRINGS Since objects contain the values you assigned to them, you can use objects in your calculations: You can store text as well by using double quotes : my_name <- stefan x+y x*y x-y We call such a piece of text a character string. x/y What happens if we compute this?: x + my_name

25 DATA TYPES Understanding the different data types in R is really important! x and my_name are vectors of a different class. You can check the type of an R object with the There are six types of atomic vectors the most basic/simple object in R: class() function: class(x) class(my_name) Logical Integer Double Numeric Most important! x = numeric my_name = character Character Complex Raw

26 COMBINING VALUES Before we continue exploring the different data types, it s useful to learn how to combine values You can combine/concatenate values by using the c() function: numbers <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10) Let s combine other data types: Character: my_full_name <- c( stefan, breet ) To see the result, type the name of the object and hit enter, or use the print() function: print(numbers) Logical: true_or_false <- c(true, FALSE, TRUE, FALSE) class(true_or_false)

27 FUNCTIONS You use functions to do most of your computations and analyses Every function consists of two parts: a name and a set of arguments (including an object): Let s try out a couple of functions: length(numbers) mean(numbers) mean(object,,, ) median(numbers) max(numbers) To see a description of the arguments, you can use the help function by putting a question mark (?) before the function min(numbers) summary(numbers) name:?mean

28 LOGICAL OPERATORS Also known as conditional statements You can compare R objects with logical operators. The result is a logical vector (i.e., TRUE / FALSE / NA) Examples: y < x y > x Operator Description y == x < less than <= less than or equal to > greater than >= greater than or equal to == exactly equal to!= not equal to x y x OR y x & y x AND y y!= x y > numbers Create a new object z: z <- 4 Check if z is equal to x or y: z == x z == y

29 LISTS A list can contain different types of data Atomic vectors can only contain data of the same type (numeric, logical or character). Lists and data frames can contain different data types. You can access elements of the list with the $ sign by calling their name: my_first_list$numbers Let s make a list of the objects we created with the list() function and provide them with a name: my_first_list <- list( my_name = my_full_name, numbers = numbers, Or by using brackets [ ]. Acces the first element as follows: my_first_list[1] And the second element: my_first_list[2] logical = true_or_false)

30 LIST VS DATAFRAME Let s take a closer look at our list Element Name Values Length my_name "stefan" "breet" 2 numbers logical TRUE FALSE TRUE FALSE 4 Note: every list element has a different length!

31 LIST VS DATAFRAME A dataframe can store different data types as well If every vector has the my_name numbers logical same length, we call "stefan" 1 TRUE the object a data "breet" 2 FALSE frame (existing of rows NA 3 TRUE and columns) and we NA 4 FALSE call the vectors variables. NA 5 NA NA 6 NA NA = Not available = NA 7 NA missing data NA 8 NA

32 SUMMARY What do you need to remember? Data Structures Operators Functions Atomic Vectors (Character, Numeric, Logical) Matrices Lists Data Frames Assignment Operator x <- value Arithmetic Operators + - / * ^ Logical Operators < <= > >= ==!= function_name( argument1, argument2, argument3, )

33 HOW TO USE R THE DATA ANALYSIS PROCESS

34 ABALONE DATA Data Analysis in R: An Example Abalones are marine snails and their meat is used for food, with prices in China reaching levels up to $ 85,- per kilogram. The dataset contains the physical measurements of 4176 abalones The research question: Can we predict the age of abalones based on their physical measurements?

35 THE DATASET Predicting the age of abalone from physical measurements Variable Name Data Type Measurement Description Sex nominal M, F, and I (infant) Length continous mm Longest shell measurement Diameter continous mm perpendicular to length Height continous mm with meat in shell Whole Weight continous grams whole abalone Shucked Weight continous grams weight of meat Viscera Weight continous grams gut weight (after bleeding) Shell Weight continous grams after being dried Rings integer +1.5 gives the age in years Source: UCI Machine Learning Respoitory

36 ALWAYS CREATE AN RSTUDIO PROJECT Creating a project makes file management easier Managing Files The RStudio Project is the place where you store raw and tidy data, the scripts you use to process and analyse the data, and output such as plots and tables or dynamic reports. Create a new project with the name R Workshop in a new directory. Loading and Saving Files A project automatically specifies the working directory, so you can use relative paths to access and save files. Absolute Path = /Users/Stefan/Documents/ data.csv or C:/Documents/data.csv Relative Path =./data.csv The dot represents the working directory. Note: You should never use absolute paths in your scripts! Click here for more information about projects

37 ALWAYS CREATE A SCRIPT Store your code in a script so you can rerun it later Use the console to experiment with code, but put it in a script as soon as you have written code Open a new R Script and save it with the name R Workshop that works and does what you want. RStudio s script editor wil also highlight syntax errors.

38 HOW TO IMPORT DATA INTO R You can import various file formats by using the appropriate package Rectangular Data Excel SPSS SAS STATA.csv /.tsv /.fwf.xls /.xlsx.sav.sas7bdat.dta base R readr readxl haven Key Functions: read.table() read.csv() read.csv2() Key Functions: read_csv() read_tsv() read_table() Key Functions: read_excel() Key Functions: read_sav() read_sas() read_dta()

39 STEP 1: DATA PROCESSING Let s import the dataset into R We re going to use the readr package to import the dataset: # Load the readr package library(readr) Copy-paste the abalone.csv dataset in the working directory # Load the dataset and assign it to an R object by giving it a name abalone <- read_csv(./abalone.csv") Note: every dataset imported, created or saved with the readr or haven packages will be stored as a special kind of data frame: the tibble. Tibbles have two advantages compared to standard data.frames: 1. Tibbles print only the first 10 rows + all the columns that fit on your screen (instead of all the rows and all the columns), so it s easier to work with large data 2. Each column reports its type (e.g., character, double, integer, etc.) If necessary, you can store a tibble as a data frame with the as.data.frame() function.

40 INSPECTING A DATA FRAME There are several ways in which you can inspect a data frame Let s see what the data looks like: print(abalone) # View the column names colnames(abalone) Notice that the column names are missing. The read_csv() function automatically reads the first row as column names. Let s change that: # Rename the columns with variable names by assigning the names to the colnames attribute colnames(abalone) <- c("sex", # Importing the dataset without column names abalone <- read_csv(./ abalone.csv, col_names = FALSE) "length", "diameter", height", whole_weight", "shucked_weight", viscera_weight","shell_weight", "rings")

41 PROCESSING DATA There are several ways in which you can inspect a data frame The first variable (sex) is stored as a character vector. For further analysis, however, we need to change it to a nominal variable with three levels (male, female, infant) with the as.factor() function: Let s check the levels of this variable with levels(): levels(abalone$sex) The levels are not in accordance with the tidy data principles, so let s rename them: # Change the class of the sex variable from character to factor abalone$sex <- as.factor(abalone$sex) # Rename the levels levels(abalone$sex) <- c("female", "infant", male") # Check the class of the variable class(abalone$sex) # Recheck the levels levels(abalone$sex)

42 DATA MANIPULATION WITH DPLYR dplyr is the best package for data manipulation The dplyr package provides a function for each basic verb of data manipulation, such as filtering, selecting columns, arranging, mutating, etc. # Select a subset of columns with select(): abalone <- select(abalone, age, sex, diameter, height, whole_weight) # Load the dplyr package library(dplyr) # Sort the dataset by age with arrange(): abalone <- arrange(abalone, age) Every dplyr functions starts with the name of the dataset as the first function argument. # Filter the dataset by age with filter(): filter(abalone, age > 25) # Create a new variable called age with mutate(): abalone <- mutate(abalone, age = rings + 1.5) # Pick the first ten rows with the slice() function: slice(abalone, 1:10)

43 DATA VISUALISATION Use ggplot to visualise your data R has built in data visualisation tools (the plot() function), but the ggplot2 package provides better quality plots and more options. # Load the ggplot2 package library(ggplot2) You can create plots in two different ways: 1. Use the quick plot function (qplot()), recommended for quick data inspection 2. Built plots layer by layer with the ggplot() function, recommended for creating highquality plots The easiest way to create plots with ggplot2 is by using the quick plot function qplot(). # Plot the dependent variable age: qplot(x = age, data = abalone) # Plot age vs diameter qplot(x = diameter, y = age, data = abalone) # Color the points by sex qplot(x = diameter, y = age, colour = sex, data = abalone)

44 THE PSYCH PACKAGE A great package for management scholars is the psych package. It offers great functions for analysing data from experiments and questionnaires, and comes with a couple of handy functions such as describe(). # Load the psych package library(psych) # Use the describe() and describeby() functions to inspect the dependent variable age describe(abalone$age) describeby(abalone$age, group = abalone$sex) The variable age is skewed and has a positive kurtosis. Let s log-transform this variable/ # Use the mutate function (dplyr) and log() function to create a new variable: abalone <- mutate(abalone, log_age = log(age)) # Inspect the dependent variable again describe(abalone$log_age) # Visualise the variable: qplot(x = log_age, data = abalone)

45 REGRESSION ANALYSIS It s really easy to conduct a regression analysis in R You can use the lm() (linear model) and glm() (generalised linear model) functions in R to conduct a regression analysis. You specify the regression formula as part of the function: # View a summary of the model: summary(model1) # Visually inspect the model plot(model1) lm(formula = age ~ sex + diameter + whole_weight, data = abalone) You can conduct moderation analysis by adapting the formula: # Store the resulting model model1 <- lm(formula = age ~ sex + diameter + whole_weight, data = abalone) model2 <- lm(formula = age ~ sex + diameter + whole_weight + diameter*whole_weight, data = abalone)

46 MODERATION ANALYSIS WITH THE PEQUOD PACKAGE The pequod package makes it easy to conduct moderation analysis The pequod package provides functions for moderated regression with residual centering, diagnostics (colinnearity), simple slopes analysis and interaction plots with the lmres() function. # Load the pequod package library(pequod) # Run model 1 model1 <- lmres(formula = age ~ sex + diameter + whole_weight, data = abalone) summary(model1) # Run model 2 model2 <- lmres(formula = age ~ sex + diameter + whole_weight + diameter*whole_weight, data = abalone) summary(model2) # Conduct a simple slopes test by indicating the predictor variable and the moderator ss <- simpleslope(object = model2, pred = diameter, mod1 = whole_weight ) print(ss) # Create an interaction plot PlotSlope(ss)

47 HOW TO USE R RESOURCES

48 A COUPLE OF USEFUL RESOURCES Books: R for Data Science (Grolemun & Wickham, 2016). Freely available online. Advanced R (Wickham, 2014). Freely available online. Online courses: Coursera EdX (e.g, Microsoft) Websites: Quick-R Stackoverflow (for asking questions)