Split-Plot Evaluation Results
The evaluation results report reflects the capability of a design relative to the specified model.
Turn on the Annotation for the report by using the View menu.
Each table in the report can be contracted to minimize scrolling.
Obtain more detailed help for individual cells by right-clicking on them.
Aliases are created when the model has more terms than the design can estimate. This can occur at build time or if runs are not performed during the experiment. No aliases found is the best result.
Defining Contrasts appears any time a fractional factorial designs is being evaluated.
Degrees of Freedom shows how the information is partitioned in the current design. There are separate sections for whole plot and subplot partitions.
Terms shows the terms in the model along with measures of correlation including power. The inputs for the calculations of these measures come from the design, the model, the power options used to determine the signal to noise ratio, and the preference setting for the significance level. The whole plot section will show lower power for the same signal to noise ratio than the subplot.
Matrix Measures is displayed if turned on in the Model options. Matrix Measures provide a way to compare a design’s capabilities to another.
Leverage contains Run number, the Leverage calculation, and Space Type (description of where the run is in space). Build Type is also shown for optimal designs. Taken together these can help diagnose and fix a bad design.
Correlations produces two correlation plots: the Correlation Matrix and “Pearson’s r”.
The Correlation Matrix is limited by the available degrees of freedom and will not include any terms aliased with terms already on the plot.
The “Pearson’s r” matrix (correlation matrix of factors) shows the relationship between the columns of the model matrix specified on the model screen.
Darker colors are more correlated. An orthogonal design will have no color in the off-diagonal cells for all the estimable terms.
The Customize menu on each matrix display has two options. The Highlight Threshold sets the value for when the numeric values are highlighted. Entering a number will change the display from the correlation color grid to a table of correlation values. Toggle Values/Colors will switch the display between the color grid, and the numeric table.
The correlation plots are useful for determining what model fitting problems may exist with a design.
For more details please see the Advanced Evaluation topic.