Fig. 4

Machine learning prediction of perceptual sensitivity (d’) and metacognitive efficiency (log M-ratio) from superiority rating scores. Relaxed elastic net regression with leave-one-sample-out cross-validation created prediction models for d’ (top row) and log M-ratio (bottom row) from superiority rating scores. Although the two models displayed similar prediction accuracy (left column), they consisted of different superiority rating items (right column). For more information, refer to Table 1. Transparent dots represent individual data points. Transparent lines represent linear regression fit using ordinary least squares. The word size was scaled relative to the (absolute value of) machine learning feature importance in the word cloud plot. Red and yellow words denote positive and negative feature importance, respectively (Table 1). R², r-squared. RMSE, root-mean-square error