A Statistical Model to Study Clustered Data: An Application in Analysis of Cerebral Microvessels

• Activity Number: 211
• Type: Contributed
• Date/Time: Tuesday, August 13, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing*
• Author(s): Mei Lu*+ and Zheng Gang Zhang+ and Hamid Soltanian-Zadeh and Nancy Oja-Tebbe and Michael Chopp
• Affiliation(s): Henry Ford Health System and Henry Ford Health Sciences Center and Henry Ford Health Sciences Center and Henry Ford Health System and Oakland University
• Keywords: Generalized Estimation Equations (GEE), ; correlated data ; Randomization test ; ), pair-wise comparison

Abstract

Burning issues in the emerging field of biostatistics for imaging studies of angiogenesis. We recently developed a quantitative three-dimensional imaging software to analyze images acquired from laser scanning confocal microcopy (LSCM) system, and this software permits us to measure in 3D vascular diameters, segments and lengths of vessels. Multiple vascular measurements are taken from two anatomic regions of each subject. We are interested in testing the vascular change over regions. Modern statistical methods can not be applied directly or may not be validated, if correlation and unpaired data are not taken into account. In this paper, we extended GEE approach to conduct the pair-wise comparisons for unpaired correlated data and had discussion on native pair-wise comparison using collapsed data. Using this new approach, we analyzed cerebral vessel difference between cortical and subcortical regions in rats (n=6). Permutation test was conducted on the same set of data; the result is consistent with this extended GEE approach.