Tools of the Trade Modeling group fMRI data

INTRODUCTION The analysis of functional magnetic resonance imaging (fMRI) data is a complicated procedure. The large data sets are computationally difficult to manipulate and special modeling techniques are necessary to deal with temporal correlation and to apply results beyond the study population. The statistical models used to analyze fMRI data require numerous steps starting with raw data and ending with an image of P-values for evaluating hypotheses. Luckily there are easy-to-use software packages that allow users to input their data and choose certain modeling options to conduct data analyses. The pitfall of the data analysis ‘black box’ is that users are often not aware why certain types of models are used and the purpose of different modeling options. The focus of this article is to describe the model used to analyze group fMRI data. The proper model for group fMRI data is the two-stage summary statistics approach of the mixed model. A mixed model is necessary to extrapolate results beyond the study sample. The two-stage summary statistics approach of this model reduces the computational burden of analyzing the large volumes of data collected in fMRI studies. We start by motivating the need for a mixed model and then go through each stage of the analysis, describing modeling options at each stage and how they impact the results.