Whole brain analysis currently faces two challenges: the increasing demand of correction for multiplicity and the usual tug of war between specificity and sensitivity. In addition, sensitivity suffers substantially because of stringent correction, while specificity is not directly considered when forming clusters. Specificity can be largely guaranteed through ROI-based analysis if ROIs can be a priori defined. Furthermore, sharing information across ROIs through an integrated model can improve model efficiency and detection power. We offer an alternative or complementary approach to the conventional methods in resolving the dilemma of multiple comparisons and dichotomous decisions. Lastly, through the approach, we promote totality and transparency in results reporting, and avoid the hard thresholding of a p-value funnel.