This is an old revision of the document!

LiRI statistics workshop: Setting appropriate contrasts in linear models

Every time a linear model is used to study the impact of a categorical predictor (or factor) on a dependent variable, it is necessary to transform this categorical predictor into numbers. R (or any other statistical software) cannot do math with words. In other words, we need to set the contrasts. Setting the contrasts right will ensure that the model answers the right questions. In this workshop, I will introduce and illustrate different ways to set contrasts and show how contrasts are directly related to research hypotheses.