Fig. 5: Statistical comparison of models at population level: TF(e) is the favored model in both FEF and SEF.
From: Integration of landmark and saccade target signals in macaque frontal cortex visual responses
Top row: the model that yields the lowest residuals (red) is used as a statistical reference, and thus gives p value of 1.0 relative to itself. Data points below p = 0.05 (the horizontal dashed line) indicate a model that gave significantly higher residuals, i.e., poor fits. a The p value statistics and comparison between different models for all FEF neurons (n = 101). In this case, TF(e) gave the best fit and all other models were significantly worse. b Same as (a) but for SEF neurons (n = 43). TF(e) is the best model overall for both sites, with all other models statistically eliminated. c Normalized mean PRESS (±SD) residuals for FEF and d SEF neurons. The values for models were normalized by dividing by the mean PRESS residuals of the best fit, i.e., TF(e). Model Definitions: TF(e): Target-relative-to-fixation in eye coordinates (note that this is the same model as ‘Te’ in our previous publications; LF(e): Landmark-relative-to-fixation in eye coordinates; TL(e): Target-relative-to-landmark in eye coordinates; T(s): Target relative to space; T(h): Target relative to head; L(s): Landmark relative to space.
