The Pareto chart is a graphical tool used to manage model selection for two-level factorial designs. We recommend using the Half-Normal Plot of Effects to choose the statistically significant effects. The primary use of the Pareto chart is to check for “one more significant effect” that was not obvious on the half-normal plot.
There are two different t limits plotted on the graph. The highest limit is based on the Bonferroni or family-wise corrected t-critical value. The lower limit is based on standard t-critical for individual effects tests. The limits are re-calculated as the terms selected for the model are changed.
The Pareto chart is best used following these steps:
1. Use the half-normal plot to select effects that are obviously larger than the others.
Use the selection tool to switch to the Pareto Chart.
2. Select the largest unselected effect on the Pareto chart to see if it belongs with error or may be a true effect.
3. Selected Effects that are above the Bonferroni Limit are almost certainly important and should be left in the model.
4. Effects that are above the t-value Limit are possibly important and should be added if they make sense to the experimenter.
5. Effects that are below the t-value limit should only be selected to support hierarchy. They can also be forced into the model by the analyst.
Click on the ANOVA tab to move on with the analysis.