Cerebral palsy (CP) is the most common cause of physical disability in childhood. Multilevel orthopaedic surgery is a key component in the clinical management of musculoskeletal pathologies in CP with the choice of surgical intervention being guided by physical exam, patient history, medical imaging and gait analysis. Clinical 3D gait analysis in particular has become an important diagnostic tool, but these methods do not provide direct objective information to assist with planning of complex surgical procedures. Contemporary approaches use anatomical musculoskeletal models that allow the calculation of muscle-tendon lengths, muscle-tendon moment arms and joint contact forces, which might improve the clinical decision-making in children with CP. Before these contemporary approaches can be used with confidence in a clinical setting it is necessary to determine whether they provide accurate and reliable kinematic results, which was the main focus of this thesis. The first study compared the joint kinematics obtained by the conventional clinical gait model with those produced by a widely used contemporary musculoskeletal model, and determined the contribution of different anatomical models, computational methods and marker sets on joint kinematic differences using anatomical models developed from magnetic resonance images (MRI). MRI and 3D motion capture data were collected from eight children with CP. Kinematic differences of up to 13° were found between the traditional and contemporary gait models, with 94% of these differences attributed to differences in the anatomical models. Conversely, the Direct Kinematics (DK) versus Inverse Kinematics (IK) computational methods and different marker sets only had a minor impact on joint kinematics.