The analysis can proceed once the response data has been entered. If more than one response has been entered, you see multiple nodes displayed on the left side on the screen under the Analysis node. Each response must be analyzed individually.
Analyze one response at a time by following these steps:
Click on the response node (name) on the left.
Choose the appropriate transformation. Otherwise, leave the option at “None.”
Choose the model:
- For factorial designs the half-normal plot of effects is the default method for choosing important model terms.
- For RSM, mixture, and combined designs the Fit Summary button displays sequential F-tests, lack-of-fit tests and other adequacy measures to suggest a starting point. Model terms are chosen on the Model tab.
Once a model is chosen, analysis of variance (ANOVA) is used to test the model as a whole and individual terms in the model. Other statistical test are computed if they are applicable along with descriptive statistics to aid with verifying the correct model has been chosen.
Inspect various diagnostic plots to statistically validate the model.
- If the model looks good, generate model graphs for interpretation.
- If there are issues with the ANOVA or diagnostics, more in depth analysis may be necessary. The appropriate transformation often fixes these issues.
For factorial designs, look at the interaction graphs or if there are not significant interactions the one factor graphs.
For RSM and mixture designs, look at the contour and 3D graphs.
After each response is analyzed, move on to multiple response optimization the graphical and numerical tools provided for this purpose.