Getting Started – R Projects and R Markdown

A typical statistical modeling project will consist of:

  1. importing data from Excel or text (.csv or .txt) files
  2. cleaning data
  3. initial exploratory plots
  4. analysis
  5. model checking
  6. generating plots
  7. generating tables
  8. writing text to describe the project, the methods, the analysis, and the interpretation of the results (plots and tables)

The best practice for reproducible research is to use as few software tools for these steps as possible. Too many research projects are not reproducible because the data were cleaned in Excel, and then different parts of the data were separately imported into a GUI statistics software for analysis, and then output from the statistics software was transcribed to Excel to make a table. And other parts of the analysis are used to create a plot in some plotting software. And then the tables and plots are pasted into Microsoft Word to create a report. Any change at any step in this process will require the researcher to remember all the downstream parts that are dependent on the change and to re-do an analysis, or a table, or a plot. Or a team member on the project making asks about how a particular variable was transformed and

R studio encourages best practices by creating a project folder that contains all project documents and implementing a version of markdown called R Markdown. An R Markdown document can explicitly link all parts of the workflow so that changes in earlier steps automatically flow into the later steps. At the completion of a project, a researcher can choose "run all" from the menu and the data are read, cleaned, analyzed, plotted, tabled, and put into a report with the text.

R vs R Studio

R is a programming language. It runs under the hood. You never see it. To use R, you need another piece of software that provides a user interface. The software we will use for this is R Studio. R Studio is a slick (very slick) graphical user interface (GUI) for developing R projects.

Open R Studio and modify the workspace preference

  1. Open R Studio
  2. Click on R Studio > Preferences to
  3. Click on General in the left menu
  4. diable "Restore .RData into workspace at startup"
  5. Click on the "Save workspace to .RData on exit" popup menu and choose "Never"

What's going on here? The workspace contains the values of all the objects created by the R code that you've run in the working R session. Nothing good comes from this. You want to start each R session with a clean slate, a blank workspace. This means that when you start a new R session, you will need to re-run all your code chunks to start where you left-off at the close of your last R session. This seems tedious but, be warned, bad things will happen if you save the workspace from the last session and re-load this at startup. Trust me. Just don't do it.

If you didn't modify the workspace preferences from the previous section, go back and do it

R Markdown in a nutshell

In this text, we will write code to analyze data using R Markdown. R markdown is a version of Markdown. Markdown is tool for creating a document containing text (like microsoft Word), images, tables, and code that can be output, or knitted, to the three modern output formats: html (web pages), pdf (reports and documents), and microsoft word (okay, this isn't modern but it is widely used).

The R Markdown, or .Rmd, document contains three components:

  1. a YAML header, which specifies formatting and styles for the knitted document
  2. the code "chunks", which are blocks of code that do something
  3. the space before and after the code chunks which contains text and any output images or tables from the code chunks.

Install R Markdown

Directions for installing R Markdown

R Markdown can output pdf files. The mechanism for this is to first create a LaTeX ("la-tek") file. LaTeX is an amazing tool for creating professional pdf documents. You do not need PDF output for this text, but I encourage you to download and install the tinytex distribution, which was created especially for R Markdown in R Studio.

The tinytex distribution is here.

Importing Packages

The R scripts you write will include functions in packages that are not included in Base R. These packages need to be downloaded from an internet server to your computer. You only need to do this once (although you have to redo it each time you update R). But, each time you start a new R session, you will need to load a package using the library() function. Now is a good time to import packages that we will use

Open R Studio and choose the menu item "Tools" > "Install Packages". In the "packages" input box, insert the names of packages to install the package. The names can be separated by spaces or commas, for example "data.table, emmeans, ggplot2". Make sure that "install dependencies" is clicked before you click "Install". Packages that we will use in this book are

  1. Import and wrangling packages
  • devtools – we use this to install packages that are not on CRAN
  • here – we use this to read from and write to the correct folder
  • janitor – we use the function clean_names from this package
  • readxl – elegant importing from microsoft Excel spreadsheets
  • data.table - we use the data.table way to wrangle data in this text.
  • magrittr – instead of nesting functions, we use the pipe operator from this package
  • stringr – we use this to wrangle character variables
  • forcats – we use this to wrangle factor variables
  1. analysis packages
  • emmeans – we use this to compute modeled means and contrasts
  • nlme – we use this for gls models
  • lme4 – we use this for linear mixed models
  • lmerTest – we use this for inference with linear mixed models
  • glmmTMB – we use this for generalized linear models
  • MASS – we will use glm.nb from this package
  • afex – we use this for classic ANOVA linear models
  • car – we use this for model checking
  • DHARMa – we use this for model checking generalized linear models
  • insight – we use this to learn about models
  1. graphing and tabling packages
  • ggplot2 – we use this for plotting
  • ggsci – we use this for the color palettes
  • ggthemes – we use this for the colorblind palette
  • ggpubr – we use this to make ggplots a bit easier
  • ggforce – we use this for improved jitter plots
  • dabestr – we use this to make several plot types
  • cowplot – we use this to combine plots
  • knitr – we use this to make kable tables
  • kableExtra – we use this to improve kable tables
  • lazyWeave – we use this for pretty p-values

Once these are installed, you don't need to do this again although there will be additional packages that you might install. You simply need to use the library() function at the start of a markdown script.

Create an R Studio Project for this textbook

  1. Create a project folder within the Documents folder (Mac OS) or My Documents folder (Windows OS). All files associated with this book will reside inside this folder. The name of the project folder should be something meaningful, such as "Applied Biostatistics" or the name of your class (for students in my Applied Biostatistics class, this folder could be named "BIO_413").
  2. Within the project folder, create new folders named
    1. "Rmd" – this is where your R markdown files are stored
    2. "R" – this is where additional R script files are stored
    3. "data" – this is where data that we download from public archives are stored
    4. "output" – this is where you will store fake data generated in this class
    5. "images" – this is where image files are stored
  3. Open R Studio and click the menu item File > New Project…
  4. Choose "Existing Directory" and navigate to your project folder
  5. Choose "Create Project"
  6. Check that a ".Rproj" file is in your project folder
  7. Download and move the file ggplot_the_model.R into the R folder.

Project folder with the .Rproj file and all main folders located at the first level of the project

Figure 2.1: Project folder with the .Rproj file and all main folders located at the first level of the project

The project directory should look like that in Figure 2.1. Importantly, the project file ("Applied Biostatistics.Rproj") and the main folders are all located at the first level within the project folder.

Bug alert If your .Rproj file is somewhere else (on the desktop, in the data folder, etc.), bad things will happen.

Working on a project, in a nutshell

  1. Wake up, brush teeth, and open the project by double-clicking the .Rproj icon. Alternatively, open R Studio and then use the File > Open Project to open the project. The name of the project will at the top-right of the R Studio window. We always want to work within an open project and the first workflow guarantees this. If we Open R Studio and then open a .Rmd file, we could be working within another project or no project at all. Bad things will happen.
  2. Run previous code chunks, in order (top to bottom). Write new code in code chunks and run. When we run code, we add R objects to the workspace. The workspace contains the values of all the objects created by the R code that has been run in the working session. When I save the .Rmd file, these values are not saved, only the text and code chunks in the R Markdown document. This is a feature, not a bug.
  3. When we are finished with the session, quit R Studio. If you get a popup window asking if you want to save the workspace, click "No". Then immediately go back to the section "Open R Studio and modify the workspace preference" above and follow the directions.

Create and setup an R Markdown document (Rmd)

  1. The top-left icon in R Studio is a little plus sign within a green circle. Click this and choose "R Markdown" from the pull-down menu.
  2. Give the file a meaningful title.
  3. Add your name in the Author text book. After the first time doing this, R Studio will default to your name.
  4. R Studio opens a demo document. Delete all text below the first code chunk, starting with the header "## R Markdown"

Modify the "setup" chunk

All R Markdown documents should start with a setup chunk that loads the packages containing R functions that are not in Base R and defines a few key R objects, such as the name of the data folder. A good practice is to load only the packages used by the chunks in the the document. For a student or researcher new to R and R studio, this can be confusing because they are unlikely to be aware of which functions belong to which package and, therefore, which packages should be loaded.

Here, I disregard this best practice and offer a general, all-purpose setup chunk for users of this text.

                                  knitr::opts_chunk$                    set(echo =                    TRUE)                                                        # wrangling packages                                                        library(here)                    # here makes a project transportable                                                        library(janitor)                    # clean_names                                                        library(readxl)                    # read excel, duh!                                                        library(data.table)                    # magical data frames                                                        library(magrittr)                    # pipes                                                        library(stringr)                    # string functions                                                        library(forcats)                    # factor functions                                                                          # analysis packages                                                        library(emmeans)                    # the workhorse for inference                                                        library(nlme)                    # gls and some lmm                                                        library(lme4)                    # linear mixed models                                                        library(lmerTest)                    # linear mixed model inference                                                        library(afex)                    # ANOVA linear models                                                        library(glmmTMB)                    # generalized linear models                                                        library(MASS)                    # negative binomial and some other functions                                                        library(car)                    # model checking and ANOVA                                                        library(DHARMa)                    # model checking                                                                          # graphing packages                                                        library(ggsci)                    # color palettes                                                        library(ggpubr)                    # publication quality plots                                                        library(ggforce)                    # better jitter                                                        library(cowplot)                    # combine plots                                                        library(knitr)                    # kable tables                                                        library(kableExtra)                    # kable_styling tables                                                                          # ggplot_the_model.R packages not loaded above                                                        library(insight)                                      library(lazyWeave)                                                        # use here from the here package                                    here                    <-                    here::here                                      # use clean_names from the janitor package                                    clean_names                    <-                    janitor::clean_names                                                        # load functions used by this text written by me                                                        # ggplot_the_model.R needs to be in the folder "R"                                                        # if you didn't download this and add to your R folder in your                                                        # project, then this line will cause an error                                    source_path                    <-                    here("R",                    "ggplot_the_model.R")                                      source(source_path)                                    data_folder                    <-                    "data"                                    image_folder                    <-                    "images"                                    output_folder                    <-                    "output"

echo = TRUE tells knitr to display the code within a code chunk when the R markdown file is knitted. knitr::opts_chunk$set(echo = TRUE) sets echo = TRUE for all code chunks in the containing .Rmd file (not all R markdown files in the project!).

here <- here::here makes sure that here uses the function from the here package and not some other package. Huh? Let's back up – R is an open source project and packages are written by independent programmers and scientists and not employees of some central company. When someone develops a package, they create functions that do stuff. Sometimes developers of different packages create functions that have the same name. There is a function name conflict if we load two packages with the same name. Our R session will use the function of the last loaded package as the function assigned to the name. If we want the name to be used with the function in the previously loaded package, then we need to either re-order the library() statements, or simply re-assign the name to the function that we want. This is what here <- here::here does. This script takes the function here from the package "here" and assigns it to the object here.

Let's play around with an R Markdown file

Create a "fake-data" chunk

  1. Create a new chunk and label it "fake-data". Insert the following R script and then click the chunk's run button
                                                      set.seed(4)                  n                    <-                    10                                    fake_data                    <-                    data.table(                                      treatment =                    rep(c("cn",                    "tr"),                    each =                    n),                                      neutrophil_count_exp1 =                    rnegbin(n*                    2,                                                        mu =                    rep(c(10,                    15),                    each =                    n),                                      theta =                    1),                                      neutrophil_count_exp2 =                    rnegbin(n*                    2,                                                        mu =                    rep(c(10,                    20),                    each =                    n),                                      theta =                    1)                  )                                      # View(fake_data)

This chunk creates fake neutrophil counts in two different experiments. The comment (#) sign before View(fake_data) "comments out" the line of code, so it is not run. View the data by highlighting View(fake_data) and choosing "Run selected line(s)" from the Run menu.

Create a "plot" chunk

  1. Create a new chunk and label it "plot". Insert the following R script and then click the chunk's run button
                                  gg_1                    <-                    ggstripchart(data =                    fake_data,                                      x =                    "treatment",                                      y =                    "neutrophil_count_exp1",                                      color =                    "treatment",                                      palette =                    "jco",                                      add =                    "mean_se",                                      legend =                    "none")                    +                                                        ylab("Neutrophil Count (Exp. 1)")                    +                                                        stat_compare_means(method =                    "t.test",                                      label.y =                    50,                                      label =                    "p.format")                    +                                                        NULL                                                      gg_2<-                    ggstripchart(data =                    fake_data,                                      x =                    "treatment",                                      y =                    "neutrophil_count_exp2",                                      color =                    "treatment",                                      palette =                    "jco",                                      add =                    "mean_se",                                      legend =                    "none")                    +                                                        ylab("Neutrophil Count (Exp 2)")                    +                                                        stat_compare_means(method =                    "t.test",                                      label.y =                    65,                                      label =                    "p.format")                    +                                                        NULL                                                                          plot_grid(gg_1, gg_2,                    labels =                    "AUTO")

Each plot shows the mean count for each group, the standard error of the mean count, and the p-value from a t-test. This statistical analysis and plot are typical of those found in experimental biology journals. This text will teach alterntatives that implement better practices.

Knit the Rmd

  1. Knit to an html file
  2. Knit to a word document
  3. If you've installed tinytex (or some other LaTeX distribution), knit to a pdf file