18 Mixed Effects Models
Overview
In this chapter, we’ll make a quick non-technical overview of fitting mixed effects models, focusing on the use of the function lme().
Objectives
Upon completion of this lesson, you should be able to:
- Recognize when a mixed effects model might be appropriate
- Be able to fit simple repeated measures models with
lme() - Be able to fit simple split-plot models with
lme()
Data and R Code Files
The R code file and data files for this lesson can be found on the Essential R - Notes on learning R page.
18.1 What is a Mixed Effects Model?
Here we’ll introduce the idea behind the mixed-effects model.
18.2 Repeated Measures Done the Wrong Way
Now we’ll begin a demonstration of the wrong way to analyze data that includes repeated measures, using the “machines” data included in the package nlme.
18.3 Repeated Measures Using Mixed Effects I
Here we’ll demonstrate the use of lme() to fit a mixed effects model - in this case a separate intercept for each worker.
18.4 Repeated Measures Using Mixed Effects II
Now we’ll try a different model, with the machine * worker interaction as a random effect. The substantial reducion in AIC and the more reduced patterning in the residuals suggests that this ia a superior model.
18.5 Split-plot Using Mixed Effects
Here we’ll revisit the split-plot experiment we analyzed using aov() in Lesson 13, this time with lme().
18.6 Using anova() to Compare Models
Here we’ll demonstrate the use of anova() to compare two models fit by lme() - note that the models must be nested and both must be fit by ML rather than REML.