Statistical approaches to neuroimaging analysis

The two most common image analysis goals are prediction and explanation. Both of these analyses build off or use the linear model, one of the most useful tools in statistics. Although the terms explanation and prediction are often used interchangeably, here prediction is used in the true sense of the word and explanation refers to studies of association. In the case of prediction, one may be interested in whether a set of variables can accurately predict a value, say future disease status. In the case of explanation, one may be interested in how variables explain a dependent variable of interest, such as an association between brain activity and age. This chapter introduces the linear model and differences between how it is used in prediction and association analyses. The goal is to support the understanding of methods in other chapters that rely on prediction or explanation.