1. R for Epi Workshop Module 1: Learning Base R
Sara Levintow, MSPH
PhD Candidate in Epidemiology
UNC Gillings School of Global Public Health

2. Outline Background on R Installation of R and RStudio
Tour of RStudio programming environment
Intro to R code syntax
Intro to R objects
Intro to functions

3. Code Along Type out or copy/paste code from the slides into R.
Run the code and check your work with the output included in slides.
Raise your hand if you have questions or get stuck!
Important note about copying and pasting quotation marks:
PowerPoint likes to make quotation marks curly: “ ”
R needs quotation marks to be straight: " "
Retype quotation marks in R to change from curly to straight!

4. 1. Background on R

5. Introducing R
Open source programming language and software environment for statistical computing and graphics
Created by Ross Ihaka and Robert Gentleman (University of Auckland, New Zealand)
Currently supported by the R Foundation for Statistical Computing (Vienna, Austria)
More info on the history of R at

6. Introducing R Freely available software, no cost to users
Wide variety of software facilities for manipulating and analyzing data
Crowd-sourcing results in state-of-the-art functionality
Highly extensible (via packages)
Easy to generate well-designed, publication-quality plots
“Environment” = fully planned and coherent system
Not an accumulation of specific, inflexible tools

7. Rising Popularity of R From “The Popularity of Data Science Software”
By Robert A. Muenchen

8. For SAS Users: Key Differences
SAS “gives you everything” vs. R “builds from the bottom”
R considered to be modern computer science language based on functions, objects, and abstraction.
SAS is older, more traditional card language.
R is constantly adding new technologies and statistical methods via packages.
Takes longer for SAS to incorporate, requires a new version roll-out.
Graphical data exploration and visualization are more powerful in R, but there is a learning curve.

9. Base R and Packages
Module 1 is an introduction to base R code, without packages.
Packages extend the capabilities of base R, developed by the community.
Contain additional functions, documentation, and sample data.
Packages are available from the Comprehensive R Archive Network (CRAN).
Modules 2 and 3 will introduce you to a set of packages known as the “tidyverse” – popular tools for data manipulation and visualization.

10. 2. Installation of R and RStudio

11. Installing R Go to https://cloud.r-project.org/
Click on download link for your operating system:
Follow instructions for download and installation.
I recommend installing the most recent version: R “Great Truth” released on 3/11/2019. You must install a version of R at least as recent as R (which dates back to 2013) in order to use RStudio.

12. Introducing RStudio Integrated development environment (IDE) for R
A layer that goes over the top of R, changing its appearance
Makes R easier to use, looks more like other interfaces you’ve used (SAS, Stata, etc.)
RStudio Desktop is free, open-source
Built-in tools: syntax highlighting, code completion, smart indentation, quick access to function definitions and R help, interactive debugger, extensive package development tools, versioning tools, many more…
More info here:

13. Installing RStudio
Go to
Click on download link for RStudio Desktop:
Click on the installer link for your operating system:
Follow instructions for download and installation.

14. 3. Tour of RStudio

15. Tour of RStudio

16. Script editor

17. Console

18. Workspace management

19. Tools for plotting, packages, help

20. Creating your first script in RStudio
Using icons at top left of Editor:
Create new R script
Open existing R script
Save current R script

21. Creating your first script in RStudio
Using icons at top left of Editor:
Create new R script
Open existing R script
Save current R script

22. Creating your first script in RStudio
Add a header at the top of your script:
##################################################
## Code from IPH Workshop: R for Epi
## Name
## Date
Save the script “workshop_code.R” to the folder on your computer you are using for this workshop (e.g., where you saved the births data).
This folder is your working directory (we’ll come back to this).

24. RStudio IDE Cheat Sheet
Excellent resource for getting started in the RStudio IDE.
Includes comprehensive list of keyboard shortcuts (for both Windows/Linux and Mac).
See here: IDE-cheatsheet.pdf

25. From Cheat Sheet:

26. 4. Intro to R Code Syntax

27. Syntax Basics Use <- or = as assignment operator (reads as “gets”)
new_variable <- code
Use # for comments (ignored when executing your code)
Case-sensitive
No command terminator required (e.g., semicolon in SAS)
In general, executing code is not sensitive to indentation, single vs. double quotes, or spacing (some exceptions).
However, important for readability to use consistent formatting.
Style guides from Google, the tidyverse

28. Syntax Basics Type the following into your script: # my first R code
x <- rnorm(1000, mean = 1.2, sd = 3)
summary(x)
hist(x)

29. Running Code
To run a line, put your cursor on the line with the code, and click Run or press Control+Enter (Command+Enter)
To run several lines, highlight the lines of code, and click Run or press Control+Enter (Command+Enter)
To run the entire script, click Source or Source with Echo

30. Running Code Run the following: What do you see in the Console?
# my first R code
x <- rnorm(1000, mean = 1.2, sd = 3)
summary(x)
hist(x)
What do you see in the Console?
Commands (after prompt < )
Results
What about in the Environment and Plots windows?

31. Editor
Environment
Plots
Console

32. Getting Help
Click on Help tab or run code ?function or use keyboard shortcut (F1).
?rnorm
?summary
?hist

33. Reading in and Saving Files
Set your working directory: The folder on your computer where you are currently working.
Read in data stored in this location.
Save files (data, plots, etc.) to this location.
getwd() #find the current working directory
setwd("~/Documents/R Workshop/")

34. Reading in and Saving Files
Read in the births dataset (from your working directory) using the read.csv() function:
births <- read.csv("births2012.csv", stringsAsFactors = F, header = T)

35. Reading in and Saving Files
There are many base R functions and package-specific functions for reading in data from other programs:
Base R: read.csv(), read.xls(), read.spss(), read.dta(), read.table(), read.delim()...
readR package: read_csv(), read_delim(), read_fwf(), read_tsv()...
haven package: read_sas(), read_dta(), read_sav()...

36. Reading in and Saving Files
You can also use the GUI for R to generate the import code for you:

37. Reading in and Saving Files
Save datasets to your working directory using similar code with “write” instead of “read”:
Base R: write.csv(), write.table()…
readR package: write_csv(), write_delim()...
haven package: write_sas(), write_dta(), write_sav()...

38. Reading in and Saving Files
Read in and save R datasets as RDS files (R-specific file type that is efficient at compressing the rectangular datasets we typically use in public health):
Base R: readRDS(), saveRDS()
readR package: read_rds(), write_rds()
births <- readRDS("births.Rds") #if already saved as RDS file
# ...
# data cleaning and recoding
saveRDS(births, file = "births_final.Rds") #save new version

39. View Data
View entire dataset in a separate window by clicking on its name in the Environment tab, or by running:
View(births)
View selected observations in the console:
head(births, n=10) # print first 10 rows
tail(births, n=5) # print last 5 rows
Dimensions of dataset
nrow(births) #number of rows (observations)
ncol(births) #number of columns (variables)

40. View Data View variable names in the console:
names(births)
Change variable names from uppercase to lowercase:
names(births) <- tolower(names(births))
To reference a specific variable in a dataset, use the dollar sign ($):
bir #start typing ‘births’....what happens?
births$ #what happens when you type this code?
births$wksgest #see this variable’s values printed to console

41. Variable Example: Weeks of Gestation
Get summary statistics on weeks of gestation (wksgest) across all observations:
summary(births$wksgest)

42. Variable Example: Weeks of Gestation
Based on the data dictionary, this variable is coded corresponding to weeks of gestation, with 99 for missing.
To omit 99s from any numeric calculations, we need to recode as R’s missing value: NA

43. Variable Example: Weeks of Gestation
Prior to recoding, you can save the originally coded version of wksgest as a new variable in births: wksgest_99
births$wksgest_99 <- births$wksgest
Then, for all observations where wksgest is equal to 99, recode as NA:
births$wksgest[births$wksgest==99] <- NA

44. Variable Example: Weeks of Gestation
Prior to recoding, you can save the originally coded version of wksgest as a new variable in births: wksgest_99
births$wksgest_99 <- births$wksgest
Then, for all observations where wksgest is equal to 99, recode as NA:
births$wksgest[births$wksgest==99] <- NA
This is our first example of indexing [using the brackets] – very powerful in R.
Rows (observations) and columns (variables) in a dataset can be referenced
using the brackets to perform functions on specific rows or columns of interest.

45. Variable Example: Weeks of Gestation
The distribution of weeks of gestation now has a maximum of 45, with missing values for observations previously coded as 99:

46. Variable Example: Weeks of Gestation
Use the table function to get a contingency table of counts for a single variable or to cross-classify two or more variables.
#simple example – frequency table for values of recoded wksgest
table(births$wksgest, useNA = "always")

47. Variable Example: Weeks of Gestation
Use the table function to get a contingency table of counts for a single variable or to cross-classify two or more variables.
#simple example – frequency table for values of recoded wksgest
table(births$wksgest, useNA = "always")
Variable values

48. Variable Example: Weeks of Gestation
Use the table function to get a contingency table of counts for a single variable or to cross-classify two or more variables.
#simple example – frequency table for values of recoded wksgest
table(births$wksgest, useNA = "always")
Counts

49. Variable Example: Weeks of Gestation
Use the table function to get a contingency table of counts for a single variable or to cross-classify two or more variables.
#simple example – frequency table for values of recoded wksgest
table(births$wksgest, useNA = "always")
Missing

50. Variable Example: Weeks of Gestation
Use the table function to compare the original variable (wksgest_99) to the recoded variable (wksgest).
# adding arguments to table function
table(births$wksgest, births$wksgest_99, useNA = "always",
dnn = c("recoded", "original"))

52. Recode 99 for other variables
Run summary() on the variables mage, visits, and mdif to check for 99s.
Just as you did for wksgest, recode 99s to NA for these variables.
For each variable, check your recoding by running summary().

53. Recoding answers
births$mage[births$mage==99] <- NA births$visits[births$visits==99] <- NA births$mdif[births$mdif==99] <- NA

54. Summary output

55. Add a New Variable: Preterm Birth
Preterm birth is defined as gestational age < 37 weeks.
Term birth is defined as gestational age ≥ 37 weeks.
# coding preterm as 1, term as 0
births$preterm <- ifelse(births$wksgest<37, 1, 0)
# check our coding: preterm should correspond to all births where wksgest < 37
table(births$preterm, births$wksgest, useNA = “always”)

56. Logical Operators
Slide credit: EPID 700 (Xiaojuan Li & Jordan Cates)

57. Subset Data
There are several ways to subset data in base R. We will return to indexing and also introduce the subset function.

58. Subset Data using Indexing
# Index individual variable: syntax is [which rows?]
births$wksgest[1:10]
# Indexing dataset: syntax is [which rows? , which columns?]
births[1:3, 1:5]

59. Subset Data using Indexing
Create a new dataset (named preterm_data) consisting only of preterm births:
# Select preterm birth observations, all variables
preterm_data <- births[births$preterm==1,]
# Select preterm birth observations, only study ID and preterm variables
preterm_data <- births[births$preterm==1, c("x","preterm")]

60. Subset Data using Subset Function
Create a new dataset (named preterm_data) consisting only of preterm births:
# First argument = data to subset; Second argument = logical expression
preterm_data <- subset(births, preterm==1)

61. Basic Summary Statistics
We have already seen the summary() function and will introduce other functions for descriptive statistics throughout today’s workshop.
# summary(): min, 25%, median, mean, 75%, max
summary(births$mage)
summary(births$mage[births$preterm==1])
summary(births$mage[births$preterm==0])

62. Basic Summary Statistics
We have already seen the summary() function and will introduce other functions for descriptive statistics throughout today’s workshop.
# what happens when you run the mean and standard deviation functions?
mean(births$mage)
sd(births$mage)

63. Basic Summary Statistics
Many functions in R force you to be explicit about missing data.
Will return NA if there are any missing values.
Specify na.rm=TRUE to first remove the missing observations before calculating the statistic of interest.
mean(births$mage, na.rm=TRUE)
sd(births$mage, na.rm=TRUE)

64. 5. Intro to R Objects

65. Data Frames and other objects in R
R is object-based: any time you use the assignment operator (<- or =), you are creating an object in R’s memory that can be subsequently referenced in your code.
births <- readRDS("births.Rds") # births is a data frame object
Data frames are the fundamental data structure in R (what you think of as a dataset).
2-dimensional data structure, where 1 row corresponds to 1 observation, and 1 column corresponds to 1 variable.

66. Data Frames and other objects in R
In addition to data frames, other types of objects are values, vectors, lists, and matrices.
n <- nrow(births) # storing value of births sample size
congen_anom <- c("anen","mnsb","cchd", "cdh", "omph","gast", "limb", "cl", "cp", "dowt", "cdit", "hypo") # character vector we’ll use later!
apgar_scores <- seq(0, 10, by = 1) # numeric vector
birth_dates <- list(min(births$dob),
births[50001:50005, "dob"],
max(births$dob),
Sys.Date()) # list holds different types of objects

67. Tips on Objects in R Object names
Cannot have spaces and are case-sensitive
Must start with an alphabetical character
Can use a period or underscore to form part of the name of an object
To list the objects created in an R session: ls()
If you save the workspace image when quitting R, all objects will be saved in the global environment.

68. Understand Data Structure
# Identify type of object class(births) class(births$cores) # Describe structure of object str(births) str(births$cores)

69. Common Data Types Logical: Boolean values (TRUE or FALSE) Numeric
as.logical(births$preterm)
Numeric
as.numeric(births$preterm)
Character
as.character(births$preterm)
Factor
as.factor(births$preterm)

70. Always check data type!
Functions often expect a certain type (e.g., numeric vs. character).
In particular, when reading in data from a different program (i.e., SAS to R), make sure that variable type is preserved.

71. Always check variable type!
# hypothetical situation where number of prenatal care visits was read in as character instead of numeric
births$visits_oops <- as.character(births$visits)
mean(births$visits_oops)
# recoding character version of visits to numeric
births$visits_num <- as.numeric(births$visits_oops)
Warnings: Alert messages that came up while your code ran.
Errors: Stop execution of your code.

72. Recode Numeric Variable to be Factor
Create factor variable: 0 visits, 1-9 visits, visits, >15 visits
# intro to cut function
births$visit_cat <- cut(births$visits,
breaks = c(0, 1, 10, 16, 99),
right = FALSE)
# table function to check your coding
table(births$visits, births$visit_cat, useNA = "always")

73. Factor Variables
Efficient way to encode both numeric and character information in a categorical variable.
A factor is stored as a vector of numeric values with a corresponding set of character values to use when the factor is displayed.
Factors represent ≥2 categories: you can specify the order of categories and the reference category.
Useful for plotting and regression models.

74. Creating factor variable for preterm
Use the function factor(): specifying the variable to recode into a factor and the corresponding levels (numeric) and labels (character).
# already created numeric preterm variable
births$preterm <- ifelse(births$wksgest<37, 1, 0)
# create preterm_f variable (factor) from preterm (numeric)
births$preterm_f <- factor(births$preterm,
levels = c(1, 0),
labels = c("Preterm", "Term"))

75. Frequency Table # easy way to check coding
table(births$wksgest, births$preterm_f, useNA = "always")
# bivariable distributions
table(births$preterm_f, births$visit_cat, useNA = "always")

76. Frequency Table # you can save any table as a table object
counts <- table(births$preterm_f, births$visit_cat) # only non-missing
# then can use the table as input
prop.table(counts, margin = 1) #get row proportions
prop.table(counts, margin = 2) #get column proportions

77. 6. Intro to Functions

78. Functions in R Functions are R objects (just like everything else!)
Nearly everything in R is done through functions.
There are many built-in functions: e.g., mean()
A strength of R is the user’s ability to write their own functions.

79. Built-in functions Numeric functions Character functions
Examples: log(x), log10(x), exp(x), round(x, digits=n)
Character functions
Examples: toupper(x), tolower(x), paste(..., sep=""), grep(pattern, x)
Try running this code: paste("Today is", date())
Statistical functions
Examples: mean(x), sd(x), quantile(x, probs), min(x), max(x), sum(x)
Specify na.rm=TRUE as a second argument if there is missing data
Probability functions
Examples: rnorm(n, m, sd), rbinom(n, size, prob), runif(n, min, max)

80. Table-making/Summarizing functions
prop.table()
summary()
describe() from the Hmisc package
CreateTableOne() from the tableone package

81. Writing your own functions
Syntax for writing functions:
myfunction <- function(arg1, arg2, ...) {
statements
return(object) }
Trying writing your own logit function:
logit <- function(p) {
log(p / (1-p))
Specify a probability of 0.5: what is the logit?
logit(p = 0.5)

82. Intro to the apply() family
Functions in base R to manipulate slices of data in a repetitive way without explicit use of loop constructs.
Also known as “functional functions” (functions taking a function as an argument).
Typically require very few lines of code.
Family members:
apply() lapply()
sapply() vapply()
mapply() rapply()
tapply() eapply()
Q: How can I use a loop to […insert task here…]?
A: Don’t. Use one of the apply functions.

83. General Idea
“For each of these things… (columns or rows in a dataframe, elements of a vector, etc.) …do this… (apply a function) …put it together… (compile the results of applying that function to each thing) …and give me the results.” (create a new object or print results to the console)

84. Our motivation: Congenital anomalies
There are 12 variables corresponding to congenital anomalies.
We would like to create a no_anomalies variable:
Coded 1 if all 12 variables = N
Coded 0 if ≥1 variable != N

85. Getting ready for apply()
# Character vector corresponding to 12 variable names congen_anom <- c("anen","mnsb","cchd", "cdh", "omph","gast", "limb", "cl", "cp", "dowt", "cdit", "hypo") # Reference those 12 variables in births, without much typing births[, congen_anom]

86. Run apply() function
births$no_anomalies <- apply(X = , MARGIN = , FUN = )

87. Run apply() function
births$no_anomalies <- apply(X = , MARGIN = , FUN = )
Data to perform function on
Apply function on
rows (1) or columns (2)
Any R function (built-in or user-defined)
Create new variable in births

88. Run apply() function
births$no_anomalies <- apply(X = births[, congen_anom], MARGIN = 1, FUN = function(x) { as.numeric(all(x == "N")) } )

89. Check your work
# check out rows coded as 0 births[births$no_anomalies==0, c(congen_anom, "no_anomalies")] # check out rows coded as 1 births[births$no_anomalies==1, c(congen_anom, "no_anomalies")]

90. Module 1 Code Summary for Births Data
## read in raw data births <- read.csv("births2012.csv", stringsAsFactors = F, header = T) ## lowercase variable names names(births) <- tolower(names(births)) ## recode 99s as NA births$wksgest[births$wksgest==99] <- NA births$mage[births$mage==99] <- NA births$visits[births$visits==99] <- NA births$mdif[births$mdif==99] <- NA ## categorical visit variable births$visit_cat <- cut(births$visits, breaks = c(0, 1, 10, 16, 99), right = FALSE) ## preterm variable births$preterm <- ifelse(births$wksgest<37, 1, 0) births$preterm_f <- factor(births$preterm, levels = c(1, 0), labels = c("Preterm", "Term")) ## congenital anomalies congen_anom <- c("anen","mnsb","cchd", "cdh", "omph","gast", "limb", "cl","cp","dowt","cdit", "hypo") births$no_anomalies <- apply(X = births[, congen_anom], MARGIN = 1, FUN = function(x){as.numeric(all(x=="N"))})

91. End of Module 1 – Preparing for Module 2
During the break, leave this code running in your RStudio session to install the ‘tidyverse’ packages (takes ~5-10 min):
install.packages('tidyverse')