Behavioral Data Analysis with R and Python

Part III. Robust Data Analysis

Ideal data is large, complete, and regularly shaped (e.g., normally distributed in the case of numeric variables). This is the data you see in introductory stats courses. Real-life data is often less accommodating, especially when dealing with behavioral data.

In Chapter 6, we’ll see how to handle missing data. While missing data is a common occurrence in data analysis, behavioral data adds a layer of complexity: which values are missing is often correlated with individual characteristics and behaviors, and that introduces bias in our analyses. Fortunately, using CDs will allow us to identify and resolve such situations as well as possible.

In Chapter 7, we’ll talk about a type of computer simulation called the Bootstrap. It’s a very versatile tool that is particularly well suited to behavioral data analysis: it allows us to appropriately measure uncertainty around our estimates when dealing with small or weirdly shaped data. Moreover, when designing and analyzing experiments, it offers an alternative to p-values that will make our lives much simpler.