1  Introduction

This book describes workflow that I’ve found to be efficient in making reproducible research reports using R with Rmarkdown and now Quarto in data analysis projects. I start with a fairly complete case study of survival patterns of passengers on the Titanic that exemplifies many of the methods presented in the book. This is followed by chapters covering importing data, creating annotated analysis files, examining extent and patterns of missing data, and running descriptive statistics on them with goals of understanding the data and their quality and completeness. Functions in the Hmisc package are used to annotate data frames and data tables with labels and units of measurement, show metadata/data dictionaries, and to produce tabular and graphical statistical summaries. Efficient and clear methods of recoding variables are given. Several examples of processing and manipulating data using the data.table package are given, including some non-trivial longitudinal data computations. General principles of data analysis are briefly surveyed and some flexible bivariate and 3-variable analysis methods are presented with emphasis on staying close to the data while avoiding highly problematic categorization of continuous independent variables. Examples of diagramming the flow of exclusion of observations from analysis, caching results, parallel processing, and simulation are presented. In the process several useful report writing methods are exemplified, including program-controlled creation of multiple report tabs.

1.1 R Code Repositories Used in This Book

This report makes heavy use of the following R packages and Github repository:

  • Hmisc package which contains functions for importing data, data annotation, summary statistics, statistical graphics, advanced table making, etc. Some new Hmisc functions are used, especially
    • addggLayers for adding extended box plots and spike histograms to ggplot2 plots, especially when run on the output of meltData
    • meltData melt a data table according to a formula, with optional substitution of variable labels for variable names
    • seqFreq for creating a factor variable with categories in descending order of sequential frequencies of conditions (as used in computing study exclusion counts)
    • hashCheck for checking if parent objects have changed so a slow analysis has to be re-run (i.e., talking control of caching)
    • runifChanged which uses hashCheck to automatically re-run an analysis if needed, otherwise to retrieve previous results efficiently
    • movStats for computing summary statistics by moving overlapping windows of a continuous variable, or simply stratified by a categorical variable
  • qreport package, a new R package available on CRAN for facilitating composition of Quarto reports, books, and web sites. Some of the qreport functions used here are
    • addCap, printCap for adding captions to a list of figures and for printing the list
    • dataChk for data checking
    • dataOverview dataset overview
    • htmlList to easily print vectors in a named list using kable
    • htmlView, htmlViewx for viewing data dictionaries/metadata in browser windows
    • kabl to make it easy to use kable and kables for making html tables
    • maketabs to automatically make multiple tabs in Quarto reports, each tab holding the output of one or more R command
    • makecolmarg to print an object in the right margin in Quarto reports
    • makecnote to print an object in a collapsible Quarto note
    • makecallout a generic Quarto callout maker called by makecolmarg, makecnote
    • makecodechunk
    • makemermaid make Quarto mermaid diagrams with insertion of variable values
    • makegraphviz does likewise for graphviz diagrams
    • scplot for putting graphs in separate chunks with captions in TOC
    • vClus for variable clustering
    • aePlot for making an interactive plotly dot chart of adverse event proportions by treatment
  • data.table package for data storage, retrieval, manipulation, munging, aggregation, merging, and reshaping
  • haven package for importing datasets from statistical packages
  • rio package for one-stop importing of a wide variety of file types
  • ggplot2 package for static graphics
  • gt package for a comprehensive and flexible approach to making tables
  • consort package for consort diagrams showing observation filtering
  • plotly package for interactive graphics
  • rms package for statistical modeling, validation, and presentation
  • knitr package for running reproducible reports, and also providing kable and kables functions for simple html table printing
  • grid and gridExtra packages for converting tables to graphs (Section 4.8)

1.2 Installing R and RStudio