About Us
BDMF Staff
The BDMF staff works closely with investigators to help develop a sound design
for their studies, determine an optimal sample size, and develop a coherent data analysis plan that reflects their study design and specific aims. Once a study proposal has been approved, the BDMF team further provides assistance in study randomization, database development, SAS programming, statistical analysis, scientific graphics, post-study data analysis, as well as writing of papers for publication and revised submission. The BDMF group also is expertly trained in advanced statistical methods including multivariate statistics (e.g., multi-dimensional scaling, factor analysis, cluster analysis, MANOVA), multi0variable regression methods (e.g., survival analysis, negative binomial regression, Poisson regression, proportional-odds model, robust regression, repeated measures analysis), non parametric modeling, and categorical data analysis. To provide the necessary functions as describe above and to accommodate future growth in the closely aligned fields of bioinformatics, genomics and functional proteomics, the BDMF is structurally organized into four functional areas.
--- Disease Etiology/Clinical Trial Design and Analysis ---
This area focuses on patient-oriented research including non-randomized studies (e.g., chart review, observational, surveillance, case-referent, case-case, outcomes, and retrospective cohort studies) and clinical trial research (e.g., two-arm single path, cross-over, group sequential, and play-the-winner clinical studies). Faculty assist principle investigators with basic study design and evaluation of potential study bias/methodologic concerns, protocol development, study questionnaire/case report form design, code book development, study randomization, implementation of group sequential stopping rules, data safety and monitoring board set-up and adverse event monitoring. They also provide assistance in conducting Meta-analysis, survival analysis, setting-up time-dependent covariate models, assessing competing risk, and analyzing correlated longitudinal data with generalized estimating equations (e.g., sandwich estimators) and mixed model methods. Additionally, BDMF analysts are on-call to provide help in setting-up cross-validated data-entry screens, FDA-compliant audit reports, interim reports, and patient tracking systems. Prior to the activation of a randomized clinical study, there are several procedures that must be followed, involving people at different levels and at different time points in the study. Central coordination is therefore essential both to prevent premature termination of the study and to diminish unnecessary work and frustrations of study personnel who usually do not have the time to assure procedures are followed correctly. In this respect, BDMF plays a pivotal role in central coordination efforts, facilitating summarization and exchange of study information.
Experimental/Laboratory research, including pre-clinical or basic science studies, requires specialized methods to analyze replicate, factorial, and other restricted randomization models often used in these studies. Correlated data collection over time also entails the use of complex repeated measures models. The BDMF team is knowledgeable in statistical methods for analyzing pre-clinical data, including specialized techniques for modeling dose-response and pharmokinetic parameters (e.g., bioavailability of compounds, apparent volume of distribution, fraction excreted unchanged, elimination rate constant, half-life elimination, mean residence time).
--- Computational Statistics/Database Management ---
The rapidly expanding fields of microarray statistical methods and computer-intensive simulation models to handle the complex-multifaceted data generated in this research. BDMF staff are familiar with linkage analysis for candidate genes, design and analysis of factorial microarray experiments, supervised learning models, gene-environment analysis techniques, neural models and other data mining methods, multivariate permutation models for multiplicity correction techniques for covariate adjustment in transmission disequilibrium tests, proteomic classification methods, hidden Markov models, k-nearest neighbor algorithms, differential gene expression profiling analysis, and pattern recognition algorithms. Assistance also is provided in the use of microarray statistical analysis software such as BioConductor, SAM, Gen MAPP, and GeneCluster, as well as parallel processing algorithms for high-speed simulation modeling.
--- Statistical Training ---
Individualized training is available to medical school investigators in database management and basic statistical methods. Further, applied short courses on medical statistics and database management are offered by the BDMF on a revolving basis.
651 Ilalo St. Biosciences Bldg. 320B
Honolulu HI 96813
Voice: (808)692-1648
Fax: (808)692-1979