This paper presents a novel approach to modeling and filtering of athetoid motion for use in assistive computer interfaces during targeting tasks such as clicking on icons. Data were recorded from a subject with athetosis during unassisted icon-clicking trials with an isometric joystick. In order to facilitate development and preliminary testing of filter designs without the cost and difficulty of repeated testing with subjects with athetosis, a quantitative model of the recorded subject data was developed using Pseudoinverse methods. Using this model within the visuomotor control loop for the icon-clicking task, a prediction filter was then developed to reduce the targetacquisition time for the user. The filter consists of an "autoregressive stretching window" system which selects five data points evenly distributed across the input and output histories to predict the intended target, together with a second-order system that smoothes the movement of the cursor. The filter has demonstrated a reduction of up to 49% in target acquisition time in preliminary experiments with the athetoid model.