Fig 1.
The relationship of the observed values (x) and predicted values (y), where fitted line ŷa = x, suggesting y and x are perfectly matched.
a) the fitted line ŷb = βx deviates from ŷa = x by (1 − β)x; and b) the fitted line ŷb = β0 + β1x deviates from ŷa = x by (1 − β1)x − β0. For each pair of predicted and observed values (i.e., (x1, y1), (x2, y2) and (x3, y3)), the green lines represent the distance between the fitted and predicted values used for calculating r, and the red dashed lines represent the distance between the predicted and observed values.
Table 1.
The mathematical definitions of relevant measures used in this study [8,15,18,30].
Fig 2.
The changes of accuracy measures (VEcv, dr and E1) and r with slope for the four simulated scenarios: a) scenario 1; b) scenario 2; c) scenario 3; and d) scenario 4.
For each scenario, a slope of 1 (blue vertical line), 0.9 (green dashed vertical line), 0.6 (blue dashed vertical line) and 0.3 (red dashed vertical line), and VEcv = 0 (blue solid horizontal line) were highlighted.
Fig 3.
The relationships between r (or r2) and VEcv for the four simulated scenarios in Fig 1 and S1 Fig.
Fig 4.
The relationships between VEcv and E1 for the fourth simulated scenario in S1 Fig and two additional extensions: 1) ε = rnorm (30, sd = 2), 2) ε = rnorm (300, sd = 2) and 3) ε = rnorm (300, sd = 3), with only the positive values presented.
Table 2.
The relation of error/accuracy measures and data properties.