Page 389 - Six Sigma Demystified

Part 3 S i x S i g m a To o l S 369

The quadratic terms are significant. Contour and/or surface plots are generated
using Stat/DOE/Response Surface/Contour Surface Plots to reveal a maxima in
the region of the last data value (B = 0; E = 14.6989).
Use Stat/DOE/Response Surface/Response Optimizer. The optimal condition is
predicted at (B = –0.51; E = 12), resulting in a predicted maximum response of
10,010.

Excel

Using Black Belt XL Add-On
Use New Chart\Designed Experiment to select the response column and analyze
results. Use the “Terms” button to specify only first-order terms for the initial
screening analysis and to define terms in each subsequent pass as described
above for Minitab.

Select Excel’s Data\Solver to use the regression function (from the DOE
analysis) to determine an optimal condition for the response. The Set Target Cell
references a cell containing the regression function, for example, “=8632 +
(–4152*B3) + (3184*B4) + (–8496*B3^2) + (–1848*B4^2) + (4377*B3*B4),” where
B3 and B4 refer to cells containing the starting value for factors B and E (respec-
tively), which are selected in the By Changing Cells field. The starting values for these
cells may be entered as the values of B and E that produce the optimal observed
response (but often can be left empty). The Min, Max, and Value of x options are used
to define the desired response as a maximum, minimum, or target.

run-Test rules

The Western Electric run-test rules add sensitivity to a control chart. These run
tests, developed by Western Electric, with some improvements by statistician
Lloyd Nelson, apply statistical tests to determine if there are any patterns or
trends in the plotted points.

When to Use

Use tests 1 through 8 on X and individual-X control charts; use tests 1 through
4 on attribute (P, U, Np, and C) charts. Run tests cannot be applied when the
underlying data or the plotted statistic is autocorrelated, such as in EWMA
charts.