Mixed models and multiple comparisons in analysis of human neurochemical maps

Examination of brain regional neurochemistry in disease states reveals differences among brain regions. Knowing where alterations in brain function are located is crucial to understanding the disease effect. The anatomical distribution of neurotransmitter receptors is now often studied using quantitative autoradiography, but the large number of brain regions involved raises serious problems for statistical analysis of such data. Due to the dependence among the subjects in case control designs, statistical analysis based on a 'mixed model' is useful. Such an analysis is illustrated using a small autoradiographic data set. The Bonferroni method, the method of Holm, and two 'False Discovery Rate'-controlling methods for adjusting P-values for multiple comparisons are compared.