Scalar-to-Image Linear Mixed Effects Models for Longitudinal Image Data

• Activity Number: 375
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Baiguo An*
• Companies: The University of North Carolina at Chapel Hill
• Keywords: Confidence interval ; High dimensional ; Hypothesis test ; Image population-specific fixed effects ; Linear mixed effects models ; Total variation

Abstract

We propose a scalar-to-image linear mixed effects model (SILMM) on scalar response with image and/or scalar population-specific fixed effects. A penalized total variation method is developed to estimate the fixed effects coefficients together with an estimation process for the variance components contained in SILMM. An upper bound for the estimation error is given. Confidence intervals and hypothesis test for high dimensional models is a very important problem. We also consider confidence intervals for coefficients of model, and propose a global hypothesis testing process for the significance of image fixed effects. Plenty of simulation studies are shown that the proposed estimation method has a superior performance against many existing method, and also demonstrate that the confidence intervals and hypothesis test method performs very well. We apply SILMM to analysis a hippocampus dataset obtained from the Alzheimers Disease Neuroimaging Initiative (ADNI) study.