Abstract
Recently, there has been increased interest in comparing the trajectories of movements made under different conditions to infer information about cognitive processes relating to aspects of motor control such as action planning. One of the more recent analysis methods involves computing the area between two trajectories to targets on opposite sides of the participant's midline for each experimental condition and then submitting those areas to a repeated measures ANOVA. Unfortunately, this method necessarily collapses the nuanced trajectory information into a single score. Therefore, we propose a new method - Bayesian Hierarchical Gaussian Process Regression (BHGPR) - which can be used to compare the entire trajectory among experimental conditions. The experimental data that was used to compare these analysis methods were taken from a study in which participants made reaching movements to targets, appearing on either side, preceded by either high (78.5%) or low predictive cues. The authors from this past study had predicted that movements to non-valid targets preceded by predictive cues would contour a lesser area than those from any other condition. The results from the comparison between traditional methods of analyzing trajectories and BHGPR indicate that BHGPR can be used to compare entire trajectories using credible intervals to demonstrate specific regions where the two trajectories differ.
