This thesis presents the intent-to-treat results of a randomized controlled trial (RCT) of the efficacy of a cycling intervention for children and adolescents with cerebral palsy (CP), called the Pediatric Endurance Development and Limb Strengthening (PEDALS) study. One hundred and twenty-nine subjects were screened by a telephone interview or prescreened during a routine clinic visit. A total of 64 qualified subjects were recruited and randomized into the cycling intervention (N=33) or the non-cycling control group (N=31). Data were collected at baseline and after three months (at follow-up) for: the time required to complete a 600 Yard Walk-Run Test, the distance walked during a 30 Second Walk Test, the Gross Motor Function Measure (GMFM-66) and a self-report and/or parent report of the subjects perception of their health related quality of life from the Pediatric Outcomes Data Collection Instrument (PODCI) and from the Pediatric Quality of Life Inventory (PedsQL™). Both questionnaires were designed to measure health related quality-of-life for subjects with disabilities. Baseline, follow-up and post-treatment changes (follow-up-baseline) were compared between groups using Student's t-tests (if a Gaussian distribution) or Wilcoxon rank sum tests (if a non-Gaussian distribution), and analysis of covariance. The analyses suggest that a cycling intervention has some effects for subjects with CP in improving their physical abilities, in activities such as walking, running and jumping as measured by the GMFM-66, but no effect on the other outcomes. These analyses helped us understand the impact of the intervention on the subjects' health related quality of life and relationships between the domains of activity and participation as defined by the World Health Organization's Model of International Classification of Function and could improve our design of better physical therapy interventions.

This thesis also presents a method for survey validation using structural equation modeling (SEM). From the PEDALS study, 102 PODCI data entries representing 3 PODCI scales were selected to demonstrate the standard procedures and interpretation of SEM model abstraction for the purpose of survey validation and recalibration. The resulting best model revealed an extra factor other than original 3 PODCI factors, and showed a statistically significant improvement in model fitting. In addition, the SEM method also provided us with a way to visualize the relationship and the interactions between the questionnaire items and the summarized scales.