Figure 8
From: An eye tracking based virtual reality system for use inside magnetic resonance imaging systems

Prediction and regression errors for different models based on measurements from 7 different poses. In each graph the horizontal axis represents the sequentially numbered target fixations and the vertical axis shows the fractional regression and prediction errors as percentages of the full screen width. The mean over the full trial for regression and prediction errors for each model are shown under each graph. The smallest mean regression error is obtained with largest model, but the smallest prediction error is obtained for the model that is quadratic in pupil coordinates and linear in eye corner coordinates.