 # Fit Summary

Clicking on the Fit Summary button starts the regression calculations to fit all of the polynomial models to the selected response.

The program calculates the effects for all model terms. It produces statistics such as p-values, lack of fit, and R-squared values for comparing the models. The fit summary output is shown on screen in a report which can also be printed and/or copied to another application such as a spreadsheet or word processor.

If a statistically significant model is detected, the program will underline and note the “Suggested” model. This becomes the default model on the Model screen.

Look for:

1. The highest order model that explains significantly more of the variation in the response (p-value small).

1. Insignificant lack of fit (p-value > 0.10).

3. Reasonable agreement between Adjusted R-squared and Predicted R-squared (within 0.2 of each other).

Note

Look for the “Design Model” at the top of the list when custom models are specified or discrete factors are used. The full-order model will often be aliased in these cases. The designed for model becomes the highest order model available for analysis.

The Whitcomb Score is used to suggest a full-order model as a starting point for the analysis. It is a combination of the Sum of Squares p-value and Lack of Fit p-values, with one score generated for each of the Adjusted and Predicted R-squared.

“M” is a function of the p-value associated with the sequential model sum of squares test.

$$M= 1,\, if\, p\leq 0.05$$

$$M= \frac{0.05}{p},\, if\, p > 0.05$$

$$M= 0,\, if\, the\, model\, is\, aliased$$

“L” is a function of the p-value associated with the lack of fit tests.

$$L=\, 1,\, if\, p\leq 0.05$$

$$L=\, \frac{p}{0.10},\, if\, p> 0.10$$

Whitcomb Score $$1=M\cdot L\cdot R^{2}predicted$$

Whitcomb Score $$2=M\cdot L\cdot R^{2}adjusted$$

Aliased models automatically get a score of 0.

Two scores (higher is better) are computed. Usually both scores agree and only one model is suggested. However, it is possible for two models to be suggested. The best model is somewhere in between the two suggested models.