Table 1 External validation parameters.

R² (residuals; serial correlation)

0.2513

R² (residuals and Yobs values)

0.3870

R²Test(100% data)

0.9604

R₀²Test(100% data)

0.9507

Classical metrics

Q²F1(100% data)

0.9989

(for 100% data)

Q²F2(100% data)

0.9493

Scaled avg.Rm²(100% data)

0.8905

Scaled DeltaRm²(100% data)

0.0315

CCC(100% data)

0.9714

R²Test(95% data)

0.9583

R0²Test(95% data)

0.9563

Classical metric

R0²Test(95% data)

0.9567

(after removing

Q²F1(95% data)

0.9996

5% data with

Q²F2(95% data)

0.9563

high residuals)

ScaledAvgRm2(95% data)

0.9363

ScaledDeltaRm2(95% data)

0.0329

CCC(95% data)

0.9786

RMSEP(100% data)

0.0862

Error-based metrics

SD(100% data)

0.0665

(for 100% data)

SE(100% data)

0.0108

MAE(100% data)

0.0559

RMSEP(95% data)

0.0504

Error-based metric

SD(95% data)

0.0239

(after removing 5% data with high residuals)

SE(95% data)

0.0040

MAE(95% data)

0.0446

MAE + 3*SD(95% data)

0.1163

NCompTest

38

Number of test set compounds, range and mean (train and test)

Train range

7.8195

TrainYMean

6.7604

Test range

2.2201

TestYMean

4.1629

Response values of test set around train mean (in %)

%Y(+/-1.0)TrainMean

2.6316

%Y(+/-1.5)TrainMean

2.6316

Threshold values utilized to judge the model predictions

(0.15*TrainingSetRange)

1.1729

(0.2*TrainingSetRange)

1.5639

(0.25*TrainingSetRange)

1.9549

Result (MAE-based criteria applied on 95% data)

Prediction quality

Good