Response distribution | Estimation method | AICc | BIC | Generalised R2 (T) | Generalised R2 (V) | RASE (T) | RASE (V) | Lambda (penalty, T) |
Normal | Standard Least Squares | −5843.0 | −5802.1 | 0.993 | 0.993 | 0.078 | 0.077 | – |
Normal | Ridge | −5843.0 | −5802.1 | 0.993 | 0.993 | 0.078 | 0.077 | 0.000 |
Normal | Lasso | −5841.8 | −5800.9 | 0.993 | 0.993 | 0.078 | 0.077 | 0.071 |
Normal | Adaptive Lasso | −5843.0 | −5802.1 | 0.993 | 0.993 | 0.078 | 0.077 | 0.000 |
Normal | Elastic Net | −5842.2 | −5801.3 | 0.993 | 0.993 | 0.078 | 0.077 | 0.050 |
Normal | Adaptive Elastic Net | −5843.0 | −5802.1 | 0.993 | 0.993 | 0.078 | 0.077 | 0.000 |
Normal | Double Lasso | −5841.8 | −5800.9 | 0.993 | 0.993 | 0.078 | 0.077 | 0.071 |
Normal | Adaptive Double Lasso | −5843.0 | −5802.1 | 0.993 | 0.993 | 0.078 | 0.077 | 0.000 |
Gamma | Maximum Likelihood | −1812.5 | −1771.5 | 0.969 | 0.968 | – | – | – |
Gamma | Ridge | −1793.6 | −1752.7 | 0.969 | 0.969 | – | – | 10.708 |
Gamma | Lasso | −1808.3 | −1767.4 | 0.969 | 0.968 | – | – | 38.550 |
Gamma | Adaptive Lasso | −1812.5 | −1771.5 | 0.969 | 0.968 | – | – | 0.000 |
Gamma | Elastic Net | −1807.2 | −1766.3 | 0.969 | 0.968 | – | – | 42.452 |
Gamma | Adaptive Elastic Net | −1812.5 | −1771.5 | 0.969 | 0.968 | – | – | 0.000 |
Gamma | Double Lasso | −1808.3 | −1767.4 | 0.969 | 0.968 | – | – | 38.550 |
Gamma | Adaptive Double Lasso | −1812.5 | −1771.5 | 0.969 | 0.968 | – | – | 0.000 |
LogNormal | Maximum Likelihood | −1355.9 | −1315.0 | 0.965 | 0.966 | – | – | – |
LogNormal | Ridge | −1339.8 | −1298.9 | 0.965 | 0.966 | – | – | 8.279 |
LogNormal | Lasso | −1353.4 | −1312.5 | 0.965 | 0.966 | – | – | 26.082 |
LogNormal | Adaptive Lasso | −1355.9 | −1315.0 | 0.965 | 0.966 | – | – | 0.000 |
LogNormal | Elastic Net | −1352.4 | −1311.5 | 0.965 | 0.966 | – | – | 30.105 |
LogNormal | Adaptive Elastic Net | −1355.9 | −1315.0 | 0.965 | 0.966 | – | – | 0.000 |
LogNormal | Double Lasso | −1353.4 | −1312.5 | 0.965 | 0.966 | – | – | 26.082 |
LogNormal | Adaptive Double Lasso | −1355.9 | −1315.0 | 0.965 | 0.966 | – | – | 0.000 |
In both training (T, n=2577) and validation (V, n=1057) sets, the models’ R2 was between 0.993 and 0.965 and RASE between 0.078 and 0.077. The lambda penalty coefficient was included, where appropriate.
AEX-FV, area under expiratory flow-volume curve; AICc, Akaike Information Criterion; BIC, Bayesian Information Criterion; RASE, root average square error.