Project R-1494

A Powerful Modeling Framework for Efficient Design and Analysis of Life Sciences Studies (Research)

A powerful encompassing biostatistical modeling framework will be developed for the design and analysis of life sciences data, leading to accelerated speed of assessment, reduced sample sizes, and more reliable conclusions for clinical and pre-clinical biopharmaceutical studies, epidemiological investigation, public health and survey research, environmental and toxicity experiments, risk assessment, biomarker, surrogate marker and surrogate endpoint evaluation, genetic studies, microarray data, and other bioinformatics experiments. The framework envisaged supersedes a variety of currently available frameworks for hierarchical and complex data structures. The methodological component capitalizes on model formulation for study design and analysis, novel sample size determination, model assessment and model diagnostics, principled treatment of incomplete data, especially when targeting human study subjects, efficient estimation and inferential methodology, and flexible and user-friendly implementation in standard software packages. The methodology is also adequate for psychometric validation (reliability, generalizability, and validity), thereby enabling the use of conventional clinical-trial data and thus obviating the need for expressly designed studies.

01 January 2009 - 31 December 2012