ALTERNATIVE STATISTICAL MODELS THAT ACCOUNT FOR CLUSTERING IN DENTAL IMPLANT FAILURE DATA

摘要: Longitudinal data analysis is a major component of public health care assessment. It important to know how treatments compare over time, diseases occurr and recurr, environmental or other exposures influence disease processes time. Investigations such topics involve the statistical time-to-event in various areas care.Long term dental assessment restorations have typically employed analyses that assume independence within patient. Dental naturally occur form clusters. The patient cluster correlated units (teeth) be evaluated. Statistical without acknowledgement within-cluster correlation can underestimate standard errors, which erroneously inflate significance level between-cluster predictors model.The purpose this thesis 1) review literature on implant data, 2) create suitable longitudinal file failure, 3) describe management methods used, 4) alternative models analyze clustered survival 5) show these used identify some patient-level site-level failure. We consider logistic regression, discrete survival, generalized estimating equations Cox model with frailty, examine associations between failure race, type, oral location implant. Models ignore clustering consistently overestimate race.