Reducing within-subject variation in chewing cycle kinematics--a statistical approach

High levels of within-subject variability have limited the use of chewing cycle kinematics in the experimental and clinical context. The purpose of this study was to validate a new strategy for reducing within-subject variability in chewing cycle kinematics, based on the 10 most representative cycles from a chewing sequence.

This prospective study included 25 young subjects, with normal class I occlusions. An optoelectronic recording system was used to track chin movements of subjects chewing gum (2.5 g). Computer programs provided estimates for duration and movement as well as the 3D coordinates of the chin point. The total output files were further processed for selection of 10 representative cycles based on standard scores for total duration, excursive ranges in the lateral, vertical and antero-posterior directions. Multilevel modelling procedures were used to test for significant differences.

There were no significant differences in cycle duration or excursions between the estimates for all cycles versus the 10 most representative cycles. Cycle shapes were very similar. There were no statistically significant differences in between-subject variances. All within-subject variances were smaller when using data from the 10 most representative cycles. The reduction of variance was approximately 33% for total duration and 75% for total 3D excursion.

The results validate the pre-processing strategy that selects the 10 most representative cycles from a sequence without altering cycle duration, excursions or shape or affecting between-subject variation but reducing within-subject variation substantially.