R-ticulate

12Some of What Lies Ahead

In this final chapter, we provide a teasing outlook on some important, widely used data analysis techniques and remedies for heteroscedasticity, whose in‐depth treatment would go beyond the scope of this introductory text. Therefore, we ask the readers to bear in mind that the illustrative presentation of these techniques inevitably falls short of being exhaustive. In Box 12.1, we summarise and exemplify the use of various important tidyverse functions (R package tidyverse, Wickham et al. 2019), some of which we have already seen in action in previous chapters. In addition, we provide a dichotomous key to the linear models covered in this textbook but we also go a bit further and point to extensions related to frequently encountered types of data, such as counts or rating scales, to give the reader a broader perspective and to avoid flawed analysis approaches (Box 12.2).

12.1 Generalised Linear Models

We will start with generalised linear models (GLMs), which represent a versatile generalisation of the linear model. GLMs relax the assumption of normally distributed errors by allowing alternative error distributions, which makes it possible to analyse discrete data (response variables) such as counts, binary data, and continuous variables with non‐normal errors (e.g. right‐skewed data characterised by a few extremely high values).

GLMs have three basic components:

1. (1) a distribution for the response variable and the model errors.
2. (2) a linear predictor ...