optima

Optimization of a predicted response value of a fitted quadratic equation.

This m-file is a canonical analysis for exploring and exploiting a fitted quadratic response surface by its stationary point.

Inputs:
b - vector of regression coefficients of the fitted quadratic equation: [intercept (1), linear (p), quadratic, (p) and cross-products (p[p-1]/2)]. It must to be enter in that strictly order. They can be coded or actual.

p - Number of independent variables (factors).

Outputs:
A complete summary of the canonical analysis given the critical points (in case you have utilized the coded factor levels, you should decode them to the actual values), predicted response at the stationary point, and the curvature and orientation of the quadratic response surface.

Tip:
To decode factor value (Xic) to actual (Xia),

Xia = 0.5*(Xil+Xih) + 0.5*(Xih-Xil)*Xic

where Xil=actual low level for factor Xi, Xih=actual high level for factor Xi.