STATISTICAL METHODS IN DENTAL RESEARCH

It is the purpose of this research project to continue to develop, adapt and apply experimental designs and methods of statistical analysis to dental and oral research. Emphasis will be placed on gingival and periodontal diseases, with attention focused initially on clinical variables and subsequently on microbiological variables. One of the major sets of analyses will be of the correlations that exist between measurements made on different sites within the same mouth. One result would be to help resolve the controversy as to whether the site or the person is the appropriate unit of analysis in experiments in periodontology. Another will be to characterize the heterogeneity that exists within the mouth by applying factor analyses and cluster analyses to site-specific data. The hypothesis will be tested that sites which are statistically similar on clinical variables will also be similar microbiologically. Another set of analyses will be of clinical data collected longitudinally. Methods of survival analysis will be applied to test hypotheses about the factors affecting the progression from gingivitis to periodontitis. The above analyses will be performed on data sets that have either already been provided (by the Boston Veterans Administration Longitudinal Study) or will be provided (by the Periodontal Research Centers at the Forsyth Dental Center, the Medical College of Virginia, and the State University of New York at Buffalo; and by other research centers including Columbia University and the University of Pennsylvania). In addition to data-analytic research, theoretical research on statistical methods applicable in dental studies will continue. Research will be conducted on nonparametric approaches to survival analysis, on group sequential designs for interim analyses of data from clinical trials, and on the effects of outliers on correlational and other multivariate analyses.