Multilevel models for repeated measures: A flexible approach for studying dental arch morphology

Traditional OLS and recently developed multilevel IGLS statistical techniques for analyzing repeated measures are compared and contrasted. Comparisons are based on arch width measurements for a mixed-longitudinal sample of 166 boys and girls. The multilevel IGLS procedures produce estimates that are more stable and meaningful than OLS estimates; sex differences are consistent for both the deciduous and permanent dentitions, maxillary widths are related to chronological age, and standard errors are consistently smaller based on the IGLS procedures. The multilevel procedures are also able to estimate variances and covariances-even when all measurements are not available for each subject-which can be used for predicting arch length. © 1993 Wiley-Liss, Inc.