The main goal of hypothesis-based functional neuroimaging is to arrive at a group decision for a set of data measured in different sessions. Hierarchical general linear model (GLM) is commonly used for this type of multilevel statistical inference problems. This study proposes a method that employs Bayesian Networks for analyzing hierarchical GLM. A major goal of the study is to put the main concepts of classical statistics, fixed-, random-, mixed-effects, into a Bayesian framework. The proposed method provides the posterior distributions for all the variables in the model. It is shown that it is possible to make generalizable inferences from a set of experimental data.