A great deal of discussion is found in recent periodontal literature as to whether the site or subject should be the unit of analysis. Utilization of the site as the unit of analysis allows site specific effects, but ignores within subject correlation. Utilization of the subject as the unit of analysis appropriately takes into account within subject correlation, but site specific factors cannot be analyzed. The purpose of this proposal is to investigate statistical methods which appropriately take into account within-subject correlations and can evaluate the effect that site and subject specific independent variables have on site specific response variables. The first two specific aims of this proposal investigate statistical methods when the response variable is continuous. Variance component models employing maximum likelihood estimation have already been used for analysis of continuous data in one periodontal clinical trial, however, further investigation into the statistical characteristics of periodontal data is needed to appropriately apply these methods to different types of periodontal clinical trials. In the first two specific aims of this application we are proposing to investigate periodontal data characteristics such as reproducibility and within subject correlation structures. The second two specific aims are intended to evaluate statistical methods which are appropriate when the response variable is dichotomous. First, statistical methods will be evaluated which convert continuous measures into dichotomous responses. Then statistical methods will be evaluated which appropriately analyze site specific dichotomous response variables. General Estimating Equations which are new to the statistical literature and have not been utilized in the analysis of site specific periodontal data will be evaluated. Available statistical packages do not contain these methods, so custom programing will have to be employed to apply these methods. Data for this proposal will be obtained from existing clinical trials, from a small reproducibility study in this proposal and from simulations. Data sets will be used to obtain estimates of data characteristics of periodontal data such as reproducibility, within subject correlation structures and frequency of positive responses of dichotomous data.
| Gunsolley, J C; Williams, D A; Schenkein, H A (1994) Variance component modeling of attachment level measurements. J Clin Periodontol 21:289-95 |
| Gunsolley, J C; Zambon, J J; Mellott, C A et al. (1994) Maintenance therapy in young adults with severe generalized periodontitis. J Periodontol 65:274-9 |
| Gunsolley, J C; Zambon, J J; Mellott, C A et al. (1994) Periodontal therapy in young adults with severe generalized periodontitis. J Periodontol 65:268-73 |
