6.1 EDA Overview

6.1.1 Analysis Development

All plotting work should be done in a single notebook that you clarify and expand over time. Your final plots for your presentation (and paper if applicable) should be stored as .png files in the results/ folder. Plots created for initial exploration do not need to be exported - they can be viewed interactively in your notebook. Make sure to make one final knit of your notebook so that all plots appear when uploaded to GitHub!

6.1.2 Initial Plotting

Like the data cleaning and descriptive statistics sections, this section will require some iteration. Initially, you should use ggplot2 to produce some draft plots that explore the distribution of your dependent variable (using a histogram) and independent variables (using a combination of histograms and bar plots of the original categorical variables and not their recoded dummies).

You should also produce bi-variate plots that show the relationship between your independent and dependent variables (using violin, ridge, and/or box and whisker plots as well as scatter plots). Finally, continuous variables should also be inspected for their normality using the appropriate visual techniques (including q-q plots).

Use these initial plots to make decisions about additional data cleaning and variable selection, and to develop some initial thoughts about the relationships between your variables.

6.1.3 Final Plotting

Once you have a finalized set of study variables, produce final versions of plots using ggplot2 and, if you would like, ggridges and ggstatsplot. All presentations and papers should contain a histogram and q-q plot of your dependent variable. You should also include plots showing the bi-variate relationship between your dependent variable and the main independent variable of interest.

These plots should be well-designed, use accessible color palettes, and should contain all of the additional features needed (e.g. title, notes, descriptive statistics, etc.).