Core C - Biostatistical and Imaging

University Of Alabama At Birmingham, Birmingham AL

Investigators

Linked publications, trials & patents

Abstract

The Biostatistics and Respiratory Imaging Core (BRIC or Core) provides the Program Project investigators with[unreadable] support in study design, sampling, data management, and statistical data analysis for publication. Complex[unreadable] modeling is best handled within the individual projects as is well documented in each project. The specific aims[unreadable] of this core are to:[unreadable] 1) Provide expertise in study design and statistical planning of research studies[unreadable] 2) Provide data management skills and resources for Center projects[unreadable] 3) Plan and conduct statistical analyses aside from complex modeling for research projects[unreadable] 4) Plan and conduct cross-project data analyses for non modeled data generated by the projects and data[unreadable] resulting from modeling within the projects.[unreadable] The statistical needs of the grant are in three primary areas: multivariate statistics, analysis of correlated data,[unreadable] and flexible regression modeling including non linear modeling. This does not preclude this Core from[unreadable] engaging in complex kinetic or dynamic modeling strategies if such help is requested by any investigator. Core[unreadable] support for each project is led by Dr. Al Bartolucci. He has over 25 years of experience as a statistical[unreadable] collaborator in cancer and environmental research and is also the developer of Bayesian methodologies for[unreadable] both clinical and environmental statistical applications. He has worked with correlated data models including[unreadable] GEE models. Also Dr. Bartolucci has managed several data coordinating centers for cancer clinical trials and[unreadable] VA Gulf War veteran studies. He is associated with personnel having expertise in the areas of high[unreadable] performance computing, database management and statistical analysis. Dr. Bartolucci has also spent[unreadable] considerable time with Dr. Postlethwait in the sample size and power considerations as well as analysis of his[unreadable] project on Biochemical Determinants (Project 1 of this proposal). This issue is discussed below for the[unreadable] upcoming grant period.[unreadable] Data management support activities include developing databases for all projects, establishing data[unreadable] dictionaries for easy reference by key project personnel, implementing automatic and manual quality control[unreadable] checks that ensures high quality data for analysis. Once a set of data has passed all quality control checks, it[unreadable] will be "signed-off on" as ready for analysis. Any changes to the data from that point on will carefully[unreadable] maintained using an audit trail procedure documenting the change. There are two key advantages in[unreadable] coordinating the data management through a single individual who is a member of the UAB Biostatistics[unreadable] Department: 1) it will be easier to merge datasets across projects, which may be done to explore similarities in[unreadable] response to various measures that are used by several projects; and 2) our personnel are experienced in[unreadable] analytic needs of the statistician that will help in database design decisions.[unreadable] A major thrust will be to support the scientific, statistical and computing needs of project 4. One of the aims of[unreadable] project 4 in relation to projects 2 and 3 is to apply the integrated model to establish nasal and lower airway[unreadable] dose-response relationships for site specific histochemical endpoints and evaluate the nose as a possible[unreadable] sentinel of lower airway effects. Although we will provide the analyses to integrate the goals of these projects,[unreadable] our statistical capabilities are outlined below referenced for all the projects. We will utilize several statistical[unreadable] techniques for this project including, but not limited to, correlation analysis, prediction modeling (both linear and[unreadable] non-linear) and testing for model validity. The general linear modeling (GLM) and non linear modeling (NLIN)[unreadable] procedures of the SAS software (Statistical Analysis System) version 9.1 are available for most applications.[unreadable]