Table 1 Classical statistical model performance summary
Predictive models | Adj \({R}^{2}\) | Effectiveness indicator | \(P\) value | Collinearity indicator |
|---|---|---|---|---|
Stepwise Regression | 0.489 | \(F\)(9,13,089) = 1393.606 | All variables P < 0.01 | \(D-W\) value = 0.788 |
Ridge Regression | 0.481 | \(F\)(8,13,090) = 1516.354 | All variables P < 0.01 | \(\text{NA}\) |
Robust Regression | 0.488 | \(F\)(9,13,089) = 1389.404 | All variables P < 0.01 | \(\text{NA}\) |
Quartile Regression (25%) | 0.275 | \(Y\)= − 0.406 | All variables, except geographical region P < 0.01 | \(\text{NA}\) |
Quartile Regression (50%) | 0.302 | \(Y\)= 0.131 | P < 0.01 | \(\text{NA}\) |
Quartile Regression (75%) | 0.320 | \(Y\)= 0.710 | P < 0.01 | \(\text{NA}\) |
PLS regression (1 principal component) | 0.525 | \(Q{h}^{2}\)= 1.000 | P < 0.05 | \(\text{PRESS}\) = 4.081 |
PLS regression (2 principal component) | 0.544 | \(Q{h}^{2}\)= −0.119 | P < 0.05 | \(\text{PRESS}\) = 4.021 |
PLS regression (3 principal component) | 0.552 | \(Q{h}^{2}\)= −0.176 | P < 0.05 | \(\text{PRESS}\) = 4.055 |
PLS regression (4 principal component) | 0.555 | \(Q{h}^{2}\)= −0.225 | P < 0.05 | \(\text{PRESS}\) = 4.149 |