Fig. 5

Coefficient intra-embedding distributions for the optimal models. For each target vector, the average of the bootstrap distribution of the coefficients of the corresponding optimal linear model (θ >= 0) was computed. The distributions of these averages are shown for two different analyses (full, bundled), for each subject (ID) and for each model coefficient. The distributions of the intercepts (coefficient 0) were omitted from the plots because they all showed a high and narrow peak around 0