Cheatsheet for R

Useful packages and their famous functions in data analysis and data management
R
tidyverse
Author

Yingying Yu

Published

December 10, 2023

Introduction

During the process of data analysis, especially in data wrangling and data visualization, we often perform similar workflows to reach our target. For example, some functions within the tidyverse are really useful to do exploratory data analysis (EDA). To fit our data to a statistical model, we need to first transform our raw data to a specific form of data frame. We might also have a preferred way to present our graphs from the result. This article summarized some of my routine commands, and this page would be keep updating for the future.

On the other hand, managing data is quite different from analyzing data, especially if you have a large clinical data set with is continually updating during the research period. In the second part of this page, I put some helpful commands for simple data management.

Data Analysis

Data Wrangling

Package tidyverse

Note

The package tidyverse included different small packages, below are what I use frequently:

Package Name Core Usage
dplyr Data manipulation
tidyr Tidy data
purrr Replace loops
stringr Convert strings
forcats Handle categorical variables

More tidyverse packages can be found on this website: https://www.tidyverse.org/packages/.

Package dplyr

When we first uploaded our data, the first instinct to do is look it up, the glimpse() function provide the number of rows and columns of the data frame, as well as the types of data of every column. The base R table() function can show the number of each kind of a variable.

Let’s use the 2020 U.S. Census data for example. ca_race contains the household income of different race/ethnicity by counties in California.

library(tidyverse)
ca_race <- readRDS("race.RDS")
glimpse(ca_race)
Rows: 58
Columns: 9
$ GEOID     <chr> "06001", "06003", "06007", "06011", "06013", "06017", "06019…
$ NAME      <chr> "Alameda", "Alpine", "Butte", "Colusa", "Contra Costa", "El …
$ whiteE    <dbl> 115050, 89583, 57606, 58780, 111149, 82988, 60299, 50559, 50…
$ blackE    <dbl> 56664, NA, 22302, 14946, 73306, 69519, 39621, 30263, 24107, …
$ nativeE   <dbl> 82661, 47321, 52734, 85208, 78815, NA, 52511, 32813, 41250, …
$ asianE    <dbl> 130236, NA, 49375, NA, 121358, 111964, 68274, 38456, 99030, …
$ islanderE <dbl> 97500, NA, 118667, NA, 90565, 58516, 45595, NA, NA, 73750, 9…
$ otherE    <dbl> 76546, NA, 48707, 67303, 72866, 69948, 45525, 42094, 41571, …
$ region    <chr> "Bay Area", "Capital Region", "Rest", "Capital Region", "Bay…
table(ca_race$region)

           Bay Area      Capital Region      Central Valley                Rest 
                  9                   9                   6                  27 
Southern California 
                  7
Tip

One of the most efficient way of using functions within the dplyr package is along with the pipe %>%. Since the function names are human-readable and easy to follow, stacking up the functions using pipe make your data wrangling process clean and concise.

The most common functions are
mutate(): adds new variables that are functions of existing variables
filter(): picks cases based on their values in a variable
group_by(): perform any operation by group
summarise(): calculate the mean, median, sd, min, max, first, last, n, n_distinct of a variable
arrange(): changes the ordering of the rows
case_when(): vectorise multiple if_else() statements
na.rm = TRUE: calculate by removing massing values

ca_race <- ca_race %>% 
  mutate(total = rowSums(select(.,whiteE, blackE, nativeE, asianE, islanderE, otherE), 
                         na.rm = TRUE)) %>%
  mutate(level = case_when(total < 50000 ~ "Less than 50000",
                           total <= 150000 ~ "50001 - 150000",
                           total <= 300000 ~ "150001 - 300000",
                           total <= 500000 ~ "300001 - 500000",
                           total > 500000 ~ "More than 500000",
                           TRUE ~ "Rest"))
sum_table <- ca_race %>%
  filter(region != "Rest") %>%      
  group_by(region) %>%
  summarize(White = mean(whiteE),
            Black = mean(blackE, na.rm = TRUE),
            Native = mean(nativeE, na.rm = TRUE),
            Asian = mean(asianE, na.rm = TRUE),
            Islander = mean(islanderE, na.rm = TRUE),
            Other = mean(otherE, na.rm = TRUE)) %>% 
  arrange(region)
sum_table
# A tibble: 4 × 7
  region                White  Black Native   Asian Islander  Other
  <chr>                 <dbl>  <dbl>  <dbl>   <dbl>    <dbl>  <dbl>
1 Bay Area            114912. 69085  84676  120988    92294  72606.
2 Capital Region       72706. 58022. 50799.  79919.   60816. 59030.
3 Central Valley       59379. 45909. 47797.  68488.   88245. 48077.
4 Southern California  77601. 59083. 64966.  95986.   82545. 60632.

More usage of dyplr can be found here

Package tidyr

Common functions:
pivot_longer() and pivot_wider(): to converts data between long and wide forms
complete(): make implicit missing values explicit
drop_na(): make explicit missing values implicit
fill(): replace missing values with next/previous value
replace_na(): replace missing values with a known value

sum_table <- sum_table %>% 
  pivot_longer(cols = c("White", "Black", "Native", "Asian", "Islander", "Other"),
               names_to = "Race",
               values_to = "Value")
sum_table
# A tibble: 24 × 3
   region         Race       Value
   <chr>          <chr>      <dbl>
 1 Bay Area       White    114912.
 2 Bay Area       Black     69085 
 3 Bay Area       Native    84676 
 4 Bay Area       Asian    120988 
 5 Bay Area       Islander  92294 
 6 Bay Area       Other     72606.
 7 Capital Region White     72706.
 8 Capital Region Black     58022.
 9 Capital Region Native    50799.
10 Capital Region Asian     79919.
# ℹ 14 more rows

More usage of tidyr can be found here

Package forcats

Common functions:
fct_reorder(): Reordering a factor by another variable
fct_infreq(): Reordering a factor by the frequency of values
fct_relevel(): Changing the order of a factor by hand
fct_lump(): Collapsing the least/most frequent values of a factor into “other”

ca_race$level <- fct_relevel(ca_race$level, c("More than 500000",
                                              "300001 - 500000",
                                              "150001 - 300000",
                                              "50001 - 150000",
                                              "Less than 50000"))
levels(ca_race$level)
[1] "More than 500000" "300001 - 500000"  "150001 - 300000"  "50001 - 150000"

More usage of forcats can be found here

Data visualization

Package ggplot2

This is most famous package to visualize data in the R community. Although ggplot2 is also part of the tidyverse package, we put it here for the sake of separate topics.

Common functions:
geom_point(): adds points to the plot
geom_line(): connects points with lines
geom_histogram(): creates histograms
geom_bar(): creates bar plots
geom_boxplot(): generates boxplots
geom_vline(): generate reference lines (horizontal, vertical, or diagonal)
geom_smooth(): adds a smoothed line to a scatterplot
facet_wrap(): facets the plot into multiple panels based on a categorical variable

sum_table %>% 
  ggplot(aes(x = Race, y = Value)) +
  facet_wrap(~ region, nrow = 2) +
  geom_bar(stat = "Identity", fill = "#98103E") +
  labs(title = "Household income race and region in California",
       subtitle = "2016-2020 American Community Survey",
       caption = "Source data from the U.S. Census Bureau",
       x = "Categories",
       y = "Values") + 
  theme_minimal()

More usage of ggplot2 can be found here

Data Management

In the actual data management process over time in a typical research study, you want to create multiple R script files. Each R script file will modify the previous version of the data file and then save a newly named version. This process allows you to easily recreate your work and to re-run your data management steps to an exact endpoint.

Review data

names(data): Show the column names
str(data): Show the structure of our data
head(data,2): Show the first 2 rows of data
tail(data,2): Show the last 2 rows of data
nrow(data): Show the number of rows
ncol(data): Show the number of columns
object.size(data): Size of the dataset in memory

In the package skimr, there is a function skim showing the structure and details of data

library(tidyverse)
library(skimr)
skim(ca_race)
Data summary
Name ca_race
Number of rows 58
Number of columns 11
_______________________
Column type frequency:
character 3
factor 1
numeric 7
________________________
Group variables None

Variable type: character

skim_variable n_missing complete_rate min max empty n_unique whitespace
GEOID 0 1 5 5 0 58 0
NAME 0 1 4 15 0 58 0
region 0 1 4 19 0 5 0

Variable type: factor

skim_variable n_missing complete_rate ordered n_unique top_counts
level 0 1 FALSE 4 300: 30, 150: 14, Mor: 7, 500: 7

Variable type: numeric

skim_variable n_missing complete_rate mean sd p0 p25 p50 p75 p100 hist
whiteE 0 1.00 73867.97 23435.16 41080 57731.75 66382.5 86247.25 146043 ▇▆▃▁▁
blackE 12 0.79 61877.83 21661.67 14946 45840.00 66883.5 78232.25 118322 ▃▇▇▆▁
nativeE 8 0.86 57756.06 20306.70 22961 41673.00 53716.0 72275.50 101023 ▅▇▅▅▃
asianE 9 0.84 86131.61 27743.71 38456 67396.00 80530.0 101563.00 155757 ▂▇▃▃▁
islanderE 28 0.52 77819.13 22980.83 26776 60479.75 79714.0 96130.25 120275 ▁▆▅▇▂
otherE 6 0.90 59723.19 13598.26 26711 49052.75 58672.0 69996.50 93750 ▁▇▇▆▁
total 0 1.00 339295.76 143907.43 52034 252112.25 353818.5 433545.00 639623 ▃▃▇▅▂

Manipulating records

distinct(): remove duplicates
dmy(): convert date strings to dates, can also be ymd() ydm(), and so on
bind_rows(): bind (append) two data together by rows
left-join(): merge two data together by columns
replace(): use it within mutate() to change specific data values
factor(): categorize the data using levels and labels
cut(): breaks the given vector into the intervals specified in “breaks”

# Bind baseline data together
bl_data <- bind_rows(bl_site1_raw, bl_site2_raw)
# Bind follow-up data together
f_data <- bind_rows(f_site1_raw, f_site2_raw)
# Join baseline and followup data
all_data <- left_join(bl_data, f_data, by=c("siteid"="siteid","hhid"="hhid","childid"="childid"))

# Add new column using mutate() for record ID with concatenated site/hh/childid
# Re-order the columns using select(). The everything() is all unspecified columns.
all_data <- all_data |> 
  mutate(record_id = paste0(siteid,hhid,childid)) |>  
  select(record_id,everything())
# create factor
all_data$siteid <- factor(all_data$siteid,
                   levels=c(1,2),
                   labels=c("JHU","UMD"))
# cut a variable
all_data$weight_cut <- cut(all_data$weight,breaks=c(0,10,Inf), labels=c("low","hi"))
# update a record
all_data <- all_data |> 
  mutate(reltns = replace(reltns, record_id == 156431, 2))

all_data[all_data$record_id == 156431,]
  record_id siteid hhid childid visitdate.x gender       dob visitdate.y reltns
1    156431    JHU  564      31   24-Sep-06      0 24-Apr-04   21-Sep-07      2
  bodtmp appeti fever cough breath weight lengt1 lengt2 lengt3 weight_cut
1   97.8      3     0     1      0   14.6   98.6   98.6   98.5         hi

Exporting data

# Write a csv file
write.csv(all_data, file="data_out.csv", row.names = FALSE)
# Save it in rda or RDS format
save(all_data, file = "tuesdata_tornados.rda")
# Save the workspace with the data
save.image(file="data01.RData")