• Brain function is organized in coordinated modes of spatio-temporal activity (functional networks) exhibiting an intrinsic baseline structure with variations under different experimental conditions. Existing approaches for uncovering such network structures typically do not explicitly model shared and differential patterns across networks, thus potentially reducing the detection power. We develop an integrative modeling approach for jointly modeling multiple brain networks across experimental conditions. The proposed Bayesian Joint Network Learning approach develops flexible priors on the edge probabilities involving a common intrinsic baseline structure and differential effects specific to individual networks. Conditional on these edge probabilities, connection strengths are modeled under a Bayesian spike and slab prior on the off-diagonal elements of the inverse covariance matrix. The model is fit under a posterior computation scheme based on Markov chain Monte Carlo. Numerical simulations illustrate that the proposed joint modeling approach has increased power to detect true differential edges while providing adequate control on false positives and achieving greater accuracy in the estimation of edge strengths compared to existing methods. An application of the method to fMRI Stroop task data provides unique insights into brain network alterations between cognitive conditions which existing graphical modeling techniques failed to reveal.
  • The modular behavior of the human brain is commonly investigated using independent component analysis (ICA) to identify spatially or temporally distinct functional networks. Investigators are commonly interested not only in the networks themselves, but in how the networks differ in the presence of clinical or demographic covariates. To date, group ICA methods do not directly incorporate these effects during the ICA decomposition. Instead, two-stage approaches are used to attempt to identify covariate effects (Calhoun, Adali, Pearlson, and Pekar, 2001; Beckmann, Mackay, Filippini, and Smith, 2009). Recently, Shi and Guo (2016) proposed a novel hierarchical covariate-adjusted ICA (hc-ICA) approach, which directly incorporates covariate information in the ICA decomposition, providing a statistical framework for estimating covariate effects and testing them for significance. In this work we introduce the Hierarchical Independent Component Analysis Toolbox, HINT, to implement hc-ICA and other hierarchical ICA techniques.
  • Identifying optimal designs for generalized linear models with a binary response can be a challenging task, especially when there are both continuous and discrete independent factors in the model. Theoretical results rarely exist for such models, and the handful that do exist come with restrictive assumptions. This paper investigates the use of particle swarm optimization (PSO) to search for locally $D$-optimal designs for generalized linear models with discrete and continuous factors and a binary outcome and demonstrates that PSO can be an effective method. We provide two real applications using PSO to identify designs for experiments with mixed factors: one to redesign an odor removal study and the second to find an optimal design for an electrostatic discharge study. In both cases we show that the $D$-efficiencies of the designs found by PSO are much better than the implemented designs. In addition, we show PSO can efficiently find $D$-optimal designs on a prototype or an irregularly shaped design space, provide insights on the existence of minimally supported optimal designs, and evaluate sensitivity of the $D$-optimal design to mis-specifications in the link function.