R cheat sheet

This cheat sheet outlines key R commands and options for statistical analysis, as taught in PHCM9795 Foundations of Biostatistics.

Generative AI statement

This work was developed with the assistance of NotebookLM

1. RStudio Setup & Basics

• Install R: Download from https://cran.r-project.org/. [Section 1.12]
• Install RStudio: Download from https://posit.co/download/rstudio-desktop/. You must install R before installing RStudio. [Section 1.12]
• RStudio Preferences: Set to not preserve workspace between sessions by deselecting “Restore .RData into workplace at startup” and choosing “Save workspace to .RData on exit: Never” (Mac: Edit > Settings, Windows: Tools > Global Options). [Section 1.12]
• Create a Project to simplify file navigation: File > New Project… > New Directory > New Project, then provide a directory name (e.g., PHCM9795) and browse for location. [Section 1.12]
• Start a new R Script: File > New File > R Script (or Ctrl+Shift+N/Command+Shift+N). Write commands here to save your work. [Section 1.12]
• Run a Script: Code > Run Region > Run All or select all the code, and use Ctrl+Enter or Command+Enter. [Section 1.12]
• Comments: Use # to add comments, which R ignores. [Section 1.12]
• Object Assignment: Use <- (e.g., x <- 42) to create objects. [Section 1.12]
• Vectors: Best thought of as columns of data. Combine values of the same type using the combine function: c() (e.g., age <- c(20, 25, 23, 29, 21, 27)). [Section 1.12]
• Data Frames: Essentially a spreadsheet of columns. [Section 1.12]
• Convert a single column to a data frame: Some packages (e.g. jmv) can only work with data frames, not single columns. Convert a single column to a data frame using as.data.frame() (e.g., age_df <- as.data.frame(age)). [Section 3.21]
• Case Sensitivity: R is case-sensitive (e.g., age, Age, AGE are all different names). [Section 1.12]

2. Packages

• Install a package: Use install.packages("package_name") (e.g., install.packages("jmv")) or Tools > Install Packages. Packages only need to be installed once for your R installation. [Section 1.12]
• Load a package: Use library(package_name) (e.g., library(jmv)). Must be done in each R session before using the package. [Section 1.12]

3. Data Handling & Manipulation

• Reading in data:
  • R data file (.rds): readRDS("file_path.rds") (e.g., pbc <- readRDS("data/activities/mod01_pbc.rds")). [Section 1.13]
  • Comma Separated Values (.csv): read.csv("file_path.csv") (e.g., sample <- read.csv("data/examples/mod02_weight_1000.csv")). [Section 2.20]
  • Excel file (.xlsx): read_excel("file_path.xlsx") from readxl package (e.g., survey <- read_excel("data/examples/mod03_health_survey.xlsx")). [Section 3.18]
• View data:
  • summary(dataframe): Quick overview of all variables. [Section 1.13]
  • skim(dataframe) (from skimr package): Provides detailed summaries and rudimentary histograms. [Section 1.13]
  • head(dataframe): Shows the first 6 rows. [Section 3.19]
  • tail(dataframe): Shows the last 6 rows. [Section 3.19]
  • subset(dataframe, condition): Lists observations meeting a condition (e.g., subset(sample, weight>200)). [Section 1.14]
• Modify data:
  • Create New Variable: dataframe$new_variable <- formula (e.g., survey$bmi <- survey$weight / (survey$height^2)). [Section 3.19]
  • Assign Meaningful Names: recommended for easy to read output. Simplest way is to create a new column: dataframe$newvariablename <- datafram$oldvariable. Spaces and punctuation can be incorporated by surrounding the new variable name in quotation marks: dataframe$"new variable name (units)" <- datafram$oldvariable.
  • Convert to Factor (Categorical Variable): dataframe$variable <- factor(dataframe$variable, levels=c(values), labels=c("labels")) (e.g., pbc$sex <- factor(pbc$sex, levels = c(1, 2), labels = c("Male", "Female"))). [Module 2, R notes]
  • Re-order Factor Levels: relevel(dataframe$factor_variable, ref="first_level") (e.g., drug$group <- relevel(drug$group, ref="Active")). Essential for correct calculation of measures of effect like RR or OR. [Section 6.13]
  • Recode Continuous to Categorical: cut(dataframe$continuous_var, breaks=c(limits), labels=c("labels")) (e.g., pbc$agegroup <- cut(pbc$age, breaks = c(0, 30, 50, 70, 100), right = FALSE, labels = c("Less than 30", "30 to less than 50", "50 to less than 70", "70 or more"))). [Section 2.21]
  • Set value to missing (NA): dataframe$variable = ifelse(condition, NA, dataframe$variable) (e.g., sample$weight_clean = ifelse(sample$weight==700.2, NA, sample$weight)). [Section 1.14]
  • Log transformation: dataframe$new_var <- log(dataframe$old_var) (e.g., hospital$ln_los <- log(hospital$los+1)). [Section 9.12]
  • Back-transform from log: exp(log_transformed_value) (e.g., exp(3.407232)). [Section 9.12]

4. Descriptive Statistics (Numerical)

• Continuous data:
  • summary(vector): Basic summary (Minimum, 1st Quartile, Median, Mean, 3rd Quartile, Maximum). [Section 1.13]
  • descriptives(data=dataframe, vars=c(var1, var2), pc=TRUE, ci=TRUE, splitBy=group_var) (from jmv): Comprehensive descriptive statistics, including percentiles, confidence interval of the mean, and can split by a categorical variable. [Section 1.13; Section 3.20]
• Categorical data:
  • One-way Frequency Table: descriptives(data=dataframe, vars=variable, freq=TRUE) (from jmv). [Module 2, R notes]
  • Two-way Frequency Table (Cross-tabulation): contTables(data=dataframe, rows=row_var, cols=col_var) (from jmv). Use pcCol=TRUE for column percents, and pcRow=TRUE for row percents. Use count=n for summarised data. [Module 2, R notes; Section 7.11]

5. Descriptive Statistics (Graphical)

• Continuous data:
  • Density plot: plot(density(dataframe$variable, na.rm=TRUE), xlab="Label", main="Title"). Can also use descriptives() function in jmv, with the dens=TRUE option (e.g. descriptives(data=dataframe, vars=varname, dens=TRUE). [Section 1.13]
  • Box plot: boxplot(dataframe$variable, xlab="Label", main="Title"). Can also use descriptives() function in jmv, with the box=TRUE option. [Section 1.13]
  • Scatter plot: plot(x=dataframe$x_var, y=dataframe$y_var, xlab="X Label", ylab="Y Label"). Add regression line: abline(lm(dataframe$y_var ~ dataframe$x_var)). Can also use scat(data=dataframe, x="X_var", y="Y_var", line="linear") from scatr package. [Section 8.13]
• Categorical data (Bar Charts):
  • One Categorical Variable: plot(dataframe$categorical_var, main="Title", ylab="Number of participants"). [Section 2.18]
  • Two Categorical Variables (from jmv package):
    • Clustered (side-by-side): contTables(data=dataframe, rows=row_var, cols=col_var, barplot=TRUE, xaxis="xcols"). [Section 2.19]
    • Stacked: contTables(data=dataframe, rows=row_var, cols=col_var, barplot=TRUE, xaxis="xcols", bartype="stack"). [Section 2.19]
    • Stacked relative frequency: contTables(data=dataframe, rows=row_var, cols=col_var, barplot=TRUE, bartype="stack", xaxis="xcols", yaxis="ypc", yaxisPc="column_pc"). [Section 2.19]
  • Two Categorical Variables (from surveymv package): surveyPlot(data=dataframe, vars="var_to_plot", group="grouping_var", type="stacked"/"grouped", freq="perc"/"count"). [Section 2.19]

6. Probability Distributions

While R has functions to calculate probabilities, external applets may be easier.

• Binomial probabilities https://homepage.stat.uiowa.edu/~mbognar/applets/bin.html
• Binomial distribution in R [Section 2.22]:
  • Probability of exactly x successes: dbinom(x=k, size=n, prob=p).
  • Probability of q or fewer successes: pbinom(q=k, size=n, prob=p).
  • Probability of more than q successes: pbinom(q=k, size=n, prob=p, lower.tail=FALSE).
• Normal probabilities https://homepage.stat.uiowa.edu/~mbognar/applets/normal.html
• Normal distribution in R [Section 3.22]:
  • Probability of q or less: pnorm(q, mean, sd).
  • Probability of more than q: pnorm(q, mean, sd, lower.tail=FALSE).

7. Confidence Intervals

• Proportions: BinomCI(x=k, n=n, method='wilson') (from DescTools package). [Section 6.11]
• Mean (individual data): descriptives(data=dataframe, vars=variable, ci=TRUE) (from jmv). [Section 3.23]
• Mean (summarised data): Use the custom ci_mean function provided in the notes, e.g. ci_mean(n=242, mean=128.4, sd=19.56, width=0.95). [Section 3.24]

8. Hypothesis Testing

• One-Sample T-test: t.test(dataframe$variable, mu=hypothesised_mean). [Section 4.12]
• Independent Samples T-test:
  • ttestIS(data=dataframe, vars=continuous_var, group=group_var, meanDiff=TRUE, ci=TRUE, welchs=TRUE) (from jmv). [Section 5.9]
  • Alternative base R: t.test(continuous_var ~ group_var, data=dataframe, var.equal=FALSE) [Section 5.9]
• Paired T-test:
  • ttestPS(data=dataframe, pairs=list(list(i1="var1", i2="var2")), meanDiff=TRUE, ci=TRUE) (from jmv). [Section 5.11]
  • Alternative base R: t.test(dataframe$var1, dataframe$var2, paired=TRUE). [Section 5.11]
• One-Sample Proportion (Binomial Test): binom.test(x=successes, n=trials, p=hypothesised_proportion). Also prop.test() for z-test. [Section 6.12]

Feedback? Contact me at phcm9795.biostatistics@unsw.edu.au