Summary

Let’s summarize what we’ve learned so far in this chapter.

• Linear regression: We saw that the linear model lm() is one of the most important and universally used models in modern statistical analysis. This single framework gives us the option to conduct what we often refer to as one-way or more ANOVA, ANCOVA, and linear regression, all with a unified and simple coding structure.

• Posthoc tests: We learned that post hoc tests for linear models are easy with the emmeans() and emtrends() functions.

• Statistical analysis is necessary: We saw that understanding significant interactions can be somewhat tricky. The goal should be to identify what’s driving the statistical nature of the interaction as well as to interpret what it means for the data. In addition to conducting post hoc analyses, one of the most revealing things to do is to simply plot the data and interpret the interaction visually.

• Use Anova() instead of anova(): We need to make sure to use the Anova() function instead of the anova() function in order to get the most conservative and accurate estimates of significance for the predictors in models.

• Use ggplot() instead of qplot(): Finally, we also learned that plotting nonlinear regression lines onto raw data is a little more complicated, but it isn’t too hard. We can use functions like predict() to calculate the curves and confidence intervals and use the ggplot() function to plot them beautifully.