HVTI ggplot graphics recipes

Version 2.1

Preface

Every figure in a CORR manuscript or presentation has been made before. The Kaplan-Meier curve with numbers at risk, the propensity-balance plot, the spaghetti plot of repeated echo measurements, the CONSORT diagram that opens the paper – we draw the same handful of figures over and over. This book collects them in one place, each with the code that produces it, so the next person who needs one starts from a working recipe instead of a blank script.

The recipes are built on the hvti* packages we maintain for this work, with ggplot2 underneath and a few specialist packages where they earn their place. Where a package gives you a helper, we use it. Where it does not (a plain density plot, a box plot), we fall back to bare ggplot2 and style it to match, so the figures look like they belong together regardless of how they were drawn.

Two ideas run through the whole book. First, a figure is built in two steps: a constructor prepares and validates the data, then plot() hands you a bare ggplot you finish with the usual +. Second, every example stands on its own. It generates its own sample data, so you can copy a chunk, run it, and see the figure before you point it at your own analysis. The aim is the one we hold for all of this team’s tooling: make the right path the easy path.

The chapters are grouped into parts. Getting Started covers the packages and conventions, and Formatting the styling layers – themes, colours, legends – you add on top of any figure. The catalog itself is three parts: time-to-event figures, distribution and count figures, and relationship and trajectory figures, one chapter per plot. Specialty plots handles the design diagrams (CONSORT, Sankey, covariate balance), a part on random forests covers model visualization, and the last parts handle tables and getting a finished figure out the door.