Fit a response (Y) on two or more predictors: the equation, which predictors matter, how much the predictors overlap (VIF), and the residual checks.
Only one predictor? Simple Regression
| # | ||||
|---|---|---|---|---|
| 1 | ||||
| 2 | ||||
| 3 | ||||
| 4 | ||||
| 5 | ||||
| 6 | ||||
| 7 | ||||
| 8 | ||||
| 9 | ||||
| 10 | ||||
| 11 | ||||
| 12 | ||||
| 13 | ||||
| 14 | ||||
| 15 | ||||
| 16 | ||||
| 17 | ||||
| 18 | ||||
| 19 | ||||
| 20 |
Multiple regression explains one result Y with several inputs at once (up to 10). Each coefficient is the effect of its input while the others are held constant, so you can see which settings really matter and predict Y for a combination of settings. The tool also warns when inputs overlap so much that their separate effects cannot be trusted.
Ordinary least squares, with exact p-values for each coefficient and for the model, R², adjusted R² and predicted R², residual checks with plots, unusual observations, and confidence and prediction intervals for a new setting. The calculation reproduces the certified values of the NIST reference data sets.
The data behind “Try an example”: surface roughness from 20 runs at different cutting speeds and material hardness.
| Result | Value |
|---|---|
| Speed | coefficient −0.0015337 · p < 0.001 |
| Hardness | coefficient −0.018489 · p < 0.001 |
| R² · R²(pred) | 98.53% · 98.03% |
| Prediction at Speed 1250, Hardness 40 | 1.908 · 95% PI 1.756 to 2.059 |
Both inputs matter: each extra 100 in speed lowers the roughness by about 0.15, and each extra unit of hardness by about 0.018, with the other held constant. The two inputs barely overlap, so their separate effects can be trusted.