Chapter 12 Re-shaping data

Thus far in this book we have worked with data sets which from the beginning were tidy. However, whether or not a data set is tidy depends on what the unit of analysis we are working on is and how we perceive what the variables are.

In this chapter, we will look a bit deeper into the concept of tidy data and how we can re-shape our data to be tidy in relation to our unit of analysis. We will do this using two verb-functions from tidyverse, pivot_longer() and pivot_wider() which are inverses of one-another (i.e. if you take a data set and pivot it wider, you can reverse the process by pivoting it longer and vice versa).

12.1 Wide data and long data

pivot_wider() and pivot_longer() refer to the process of taking either wide or long data and making it either long or wide. But what is wide and long data? In short, wide data refers to data sets which units (not necessarily units of analysis) such as countries, individuals, etc have one row each while long data refers to data sets in which units have multiple rows each, for instance one column for the case, one for the variable, and one for the value.

Using these definitions, it seems obvious that the wide data, in which each row corresponds to a single unit, would always be the tidy alternative. However, as we will see which data is tidy depends on the analysis you are doing, and primarily which the unit of analysis for the analysis is.

Let’s look at two simple examples. Here you can download two data files containing variables of population and gdp on 163 countries. Let’s read these into R and have a look: