2.2 Exploratory Plots vs. Expository Plots
Data visualizations can be made either for exploratory or expository purposes. To frame the difference between the two, let’s take a look at the roots of these words.
Exploratory comes from “explore”, meaning “to look into closely; scrutinize; examine.”1 Exploratory plots are the ones that a modeler often makes when she seeks to better understand a dataset and its variables. When building such plots, she may be looking to gain insights about variable relationships, identify outliers, and learn about things like concentration and skew among the individual variables.
With exploratory plotting, there are no real “rules” that one must follow. Typically, these are plots that will only be seen by the modeler, or the modeling team, that is analyzing the data. Therefore, details such as axis labels, descriptive titles, on-plot text annotations, etc. are not really important.
Expository, on the other hand, comes from “expose”, meaning “to present to view; exhibit”; display.”2 Expository plots are built for some audience other than the maker of the graph – this could be another department within the firm, an external client (if you are working at a consulting company), or it could be an even wider audience, if you are writing a paper or otherwise performing research with the aim of publication.
When making expository plots, your graphs should be completely understandable to an outside reader, without requiring any additional interpretation or explanation beyond what is shown in the plot itself, plus your axes, your title, and any annotations.
To see why this distinction can be important, take a moment to look at the plot below. Before you do, assume only that there are 21 homes in the Elm Street Neighborhood, and that radon is a gas that can be dangerous to a home’s occupants if present at high levels.
Now that you have reviewed the plot, what does it show? Does “1” here mean that the home was inspected, and found to be okay? Does “1” mean that the home was inspected, and found to have dangerous levels of radon? Or is there simply no way to really know this, based on the way the graph is presented?
As an exploratory plot, there would be nothing wrong with this barplot; presumably, the modeler would have access to this information, and would understand the distinction between the “0” and “1” groups. However, there is nothing obvious or intuitive about this to an outside audience.
Now, view the plot below. Note how the simple inclusion of descriptive axis labels completely changes a reader’s ability to understand what it shows.
1 https://www.dictionary.com/browse/explore
2 https://www.dictionary.com/browse/expose