Figure 3.

Evaluating the Performance of Different Feature Spaces

(A) Raw test set performances in the left and right auditory cortex (AC) for models based on different feature spaces shown on the horizontal axis. See Table 1 for an explanation of the feature spaces and their shorthand notations. Each color codes for a single participant (n = 24); each dot is one test set. Pooled medians are indicated with black lines.

(B) Samples from the posterior distribution of differences of beta estimates (competing feature spaces minus benchmark Sg & Art feature space, results left of the red line thus reflect that the Sg & Art feature space has a higher performance; results right of the red line indicate that the competing feature space has a higher performance). Feature spaces are color coded as indicated.

(C) Percentage of samples in favor of hypotheses of differences of beta estimates between all feature spaces. Hypotheses are color coded according to the same color mapping as in (B), which corresponds to the bottom row and right column of the matrix shown here.

(D) Samples from posterior distribution of differences of beta estimates of individual participant’s right ACs minus left ACs. Color mapping is the same as in (A).

See also Figures S2 and S3.