Additional Comments about choosing effects:
When choosing effects from a probability plot, remember to choose only those effects that are clearly off the guide line after it is moved to represent the noise. True, detectable effects are larger than noise, and are part of a different distribution (below and right of the guideline) .
Shapiro-Wilk Normality Test:
Stat-Ease performs a Shapiro-Wilk hypothesis test on the normality of the unselected terms on the Effects plot. The null hypothesis is that the unselected terms represent noise. Noise is assumed to follow a normal distribution. When the selection of statistically significant terms is complete, the Shapiro-Wilk p-value should above 0.10 to indicate there is no significant deviation from the assumption of normality for the non-selected effects.
Note
The Shapiro-Wilk test for normality is not shown on the Residuals Normal Probability plot because this plot violates the assumption of independence by ordering the residuals. Therefore, it is not an appropriate test.