Applications of Survey Sampling Methodology to Analysis of Repeated Measures Data Structures in Dentistry
Glenn M. Davies
and Gary G. Koch
1993, Presented
at the Biometrics Society Spring Meetings
Keywords:
Intra-patient correlation, Survey data regression, Stratified
cluster sample, Dental data
Abstract
In the dental
sciences, researchers often encounter data which are observed at the
site or tooth level within each patient. Consequently, statistical
methods for repeated measures or clustered outcomes are needed to
analyze dental data with this structure. Survey sampling researchers
have developed regression methods to analyze continuous or
categorical data from stratified multistage cluster samples.
Statistical packages like SUDAAN (Survey Data Analysis) are available
to implement such analyses. This paper discusses applications of
SUDAAN to a cross-sectional study of school aged children and a
longitudinal study of elderly patients over 65. Continuous responses
are analyzed with SUDAAN procedures for comparing ratio means and for
multiple linear regression; dichotomous responses are analyzed with
SUDAAN procedures for logistic regression. For each application,
patients are managed as the primary sampling unit through which
adjustment for the correlated data structure within patient is achieved.