Identifying outliers

There are different methods used to detect outliers depending on whether you are analyzing one variable at a time (univariate analysis) or multiple variables at once (multivariate analysis). In the univariate case, the analysis is fairly straightforward. The multivariate case, however, is more complex. Let's examine them in detail.

Univariate outliers

One of the most direct and widely used ways to identify outliers for a single variable is to make use of boxplots, which you learned about in Chapter 11, Adding Statistics Insights: Associations. Some of the key points of a boxplot are the interquartile range (IQR), defined as the distance from the first quartile (Q1) to the third quartile (Q3), the lower whisker (Q1 - 1.5 x IQR), and the upper whisker (Q3 + 1.5 x IQR):

Figure 12.2 – Boxplot's main characteristics

Specifically, all observations that are before the lower whisker and after the upper whisker are identified...