1. Data manipulation in R: dplyr
EPID 799C
Wednesday, Sept. 27, 2017

2. Today’s Outline Key functions of dplyr
Review of coding for key variables in births dataset
Practice dplyr coding with births dataset

3. Key Functions select() filter() arrange() summarise() mutate()
For more, see this great resource: Data Wrangling Cheat Sheet
select()
Picks variables (columns) based on their names.
filter()
Picks observations (rows) based on their values.
arrange()
Changes the ordering of the rows based on their values.
summarise()
Reduces multiple values down to a single summary value.
mutate()
Adds new variables that are functions of existing variables.
group_by()
Performs data operations on groups that are defined by variables.

4. Key Functions select() filter() arrange() summarise() mutate()
For more, see this great resource: Data Wrangling Cheat Sheet
select()
Picks variables (columns) based on their names.
filter()
Picks observations (rows) based on their values.
arrange()
Changes the ordering of the rows based on their values.
summarise()
Reduces multiple values down to a single summary value.
mutate()
Adds new variables that are functions of existing variables.
group_by()
Performs data operations on groups
that are defined by variables.

5. Key Functions select() filter() arrange() summarise() mutate()
For more, see this great resource: Data Wrangling Cheat Sheet
select()
Picks variables (columns) based on their names
filter()
Picks observations (rows) based on their values
arrange()
Changes the ordering of the rows
summarise()
Reduces multiple values down to a single summary value
mutate()
Adds new variables that are functions of existing variables
group_by()
Performs data operations on groups that are defined by variables
The pipe operator %>% enables you to pass the output from one function to the input of the next function

6. Basic structure of dplyr code
Dataset %>% Select rows or columns to manipulate %>% Arrange or group the data %>% Calculate statistics or new variables of interest

7. Basic structure of dplyr code
summary <- Dataset %>% Select rows or columns to manipulate %>% Arrange or group the data %>% Calculate statistics, new variables
Creates a new object named summary that stores the output
Otherwise, output is just printed in the console

8. Manipulating the births dataset with dplyr
Key variables of interest in today’s examples
Preterm birth
Early prenatal care
Maternal age
Smoking during pregnancy
Race/ethnicity
County of residence
Review of coding for each variable

9. Preterm Birth
births$wksgest[births$wksgest==99] <- NA births$preterm <- ifelse(births$wksgest<37,1,0) births$preterm_f <- factor(births$preterm, levels = c(1,0), labels = c("preterm", "term"))

10. Early Prenatal Care
births$mdif[births$mdif==99] <- NA births$pnc5 <- ifelse(births$mdif<=5,1,0) births$pnc5_f <- factor(births$pnc5, levels = c(1,0), labels = c("Early prenatal care", "No early care"))

11. Maternal Age
births$mage[births$mage==99] <- NA

12. Maternal Smoking during Pregnancy
births$cigdur[births$cigdur=="U"] <- NA births$smoke <- ifelse(births$cigdur=="Y",1,0) births$smoke_f <- factor(births$smoke, levels = c(1,0), labels = c("Smoker", "Nonsmoker"))

13. Format Helper
CSV file with labels for the levels of the race/ethnicity and county variables to save you from typing them out
Save file to your computer and then read into R Studio
formatter <- read.csv(“birth-format-helper-2012.csv”,
stringsAsFactors = F)

14. Maternal Race/Ethnicity
births$race_f <- factor(births$mrace, levels = 1:4, labels = formatter[formatter$variable=="mrace",]$recode, ordered = T) births$raceeth <- ifelse(births$mrace == 1 & births$methnic == "N", "WnH", ifelse(births$mrace == 1 & births$methnic == "Y", "WH", ifelse(births$mrace==2, "AA", ifelse(births$mrace==3, "AI/AN", "Other")))) births$raceeth_f <- factor(births$raceeth, levels=c("WnH", "AA", "WH", "AI/AN", "Other"))

15. County of Residence in NC
births$county <- factor(births$cores, levels = formatter[formatter$variable=="cores",]$code, labels = formatter[formatter$variable=="cores",]$recode, ordered = T)

16. Practice Problems using dplyr
1) Calculate the numbers of births by early prenatal care (received early care vs. no early care). Exclude the births with missing values for prenatal care or preterm.
Pseudo-code:
use the births dataset %>%
exclude births with missing pnc5_f or preterm %>%
group births by pnc5_f %>%
summarize number of births in each group
Syntax for summary statistics is to name the new variable and set equal to the function of interest:
summarise(number = n())
Name you choose
The function n() counts the number of observations (no arguments)

17. Practice Problems using dplyr
2) Calculate the numbers of births and average age of mothers in those same groups (received early care vs. no early care).
Pseudo-code:
use the births dataset %>%
exclude births with missing pnc5_f or preterm %>%
group births by pnc5_f %>%
summarize number of births, average age, in each group
Are there mothers with missing age?
Within the mean function, include an argument to remove missing values prior to calculating the average age.

18. Practice Problems using dplyr
3) In addition to the numbers of births and average age of mothers by early care, calculate the number and percentage of preterm births in these two groups.
Pseudo-code:
use the births dataset %>%
exclude births with missing pnc5_f or preterm %>%
group births by pnc5_f %>%
summarize total births, average age, number of preterm, % preterm
Different ways to calculate % preterm:
Within summarise(): using the function mean(preterm) OR sum(preterm)/n()
Within mutate(): as a function of the variables you created in summarise() for numbers of preterm and total births

19. Practice Problems using dplyr
4) Continuing to build on your code you’ve already written, calculate the percentage of smokers in the same two groups (early care vs. no early care).
Pseudo-code:
use the births dataset %>%
group births by pnc5_f %>%
summarize total births, average age, # preterm, # smokers, % preterm, % smokers
You could try out different methods for calculating % smokers and % preterm: within summarise (using built-in functions)
within mutate (using the new variables you’ve created)
**Note: should you use the numeric variables (smoke, preterm) or factor variables (smoke_f, preterm_f) for calculating these summary statistics?**

20. Practice Problems using dplyr
5) Onto a new example: Calculate the total births in each maternal race/ethnicity group (WnH, AA, WH, AI/AN, Other). For each group, also calculate the prevalence of early care and prevalence of preterm birth.
Pseudo-code:
use the births dataset %>%
group births by raceeth_f %>%
summarize the numbers of total births, births with early care, and preterm births in each group %>%
calculate percentage of early care, percentage of preterm in each group

21. Practice Problems using dplyr
6) Final example: Calculate the prevalence of early prenatal care and prevalence of preterm birth by NC county of residence
Pseudo-code:
use the births dataset %>%
group births by county %>%
summarize the number of births with early care, number of preterm births in each county %>%
calculate prevalence of early care and preterm in each county
This output is large – how would you store it instead of just printing to the console?