实验设计文件.ppt

* * * * * * * * * * Minitab path STAT DOE Factorial Analyze Factorial Design * Response Enter Rating Terms Selected Terms A : Temp B : Time C : Chip AB AC BC ABC Click OK * * Review the Minitab session window Effect (also called a contrast)= Ave (+1) - Ave (-1) Coefficient = Effect/2 when each factor is tested at 2 levels. Why don’t we have any F or P values in the session window? Why don’t we have any detail regarding residual error? Answer: Because we have no degrees of freedom left over for the error term. The error term is used to calculate the F and p values. F = MS factor or interaction/MS error Degrees of Freedom (DF) = The amount of independent information available to calculate the Sum of Squares. DF can be severely limited through small number of run combinations in the experimental design. To compute the F and p values we must reduce the model by removing at least one selected term from the model and giving it to the error term. We should choose the interaction with the smallest effect first. * Note: If the Pareto bar extends to the right of the red dotted line it is a significant factor at the noted alpha level of 0.10. Factor A (Temp) is a significant main effect factor. The interaction between factors A and C (Temperature * Chip) is also significant interaction. The X-axis is absolute Effect (all Effects are given in absolute value). This should match the absolute value of the Effects we calculated for each factor and interaction for the Cookie Rating example. Note: if there is no red line displayed, then either no factor or interaction are significantly affecting the response variable (no factor or interaction has a p-value .10). Or, all factors and interactions are significant. This is why it is important to run both the pareto and normal probability plots as the red line can either reside directly on the right or left hand extremes of the pareto diagram. * * Note: If the point on the Normal p