Beyond GLMs: an Introduction to Mixed Models

Question 1

Key Assumptions

Answer 1

Independence of observations - dependent variables are independent of one another
Homoskedasticity - variance of the residuals should normally be constant across different levels of the independent variable(s)

Question 2

Linear Model

Answer 2

accuracy = Bo + B1 * RTi + Ei

Question 3

Wilkinson Notation

Answer 3

accuracy ~ 1 + RT

Question 4

‘Simpson’s Paradox

Answer 4

Hierarchical approach

Question 5

Mixed Effects Models

Answer 5

Some effects are fixed, others may vary randomly
Number of random effects varies depending on how many sources of variability there are
More specific in the effect and allows the collection of variability within the data

Question 6

Summary

Answer 6

Data may have a hierarchical structure - may manipulate a variable within-participant, and then want to compare how the effect of this manipulation varies between
Some of our observations are now non-independent - have arisen from the same participant
Mixed effects models provide a general way to account for hierarchical structure