Advanced

Mixture Options

Include Linear+Squared terms, enable partial quadratic mixture (PQM) model: Check this box to enable more advanced mixture modeling and analysis.

Show orthogonal direction: Enables the orthogonal direction for trace plots. The default display only shows Cox and Piepel directions which are better for interpreting the shape of the mixture response surface. Orthogonal direction for a component X is perpendicular to simplex that does not include X.

Propagation of Error (POE) Options

Use POE adjustment for intervals and predictions: Uncheck this box to NOT include the error propagated from factor variation in the interval estimates. This will be the same a making all the factor variation 0.

Use 2nd Order POE (process factor designs only): Uncheck this box to make the output match an earlier version. The 2nd order approximation for propagated error is more accurate than the first order approximation used in earlier versions of the software.

Correct biased predictions of mean for transformations: Models needing a transformation predict the median rather than the mean when untransformed. With this correction on, unbiased predictions for the mean are provided on the point prediction node. Confidence interval endpoints are also corrected.

Use adjusted model for predictions when curvature present: For factorial designs with center points, special calculations are done to separate the curvature from the lack of fit test. This generates two ANOVAs: the (default) unadjusted (curvature ignored) where all the response data is used to estimate the terms in the factorial model, and the adjusted model (curvature estimated) where the center points are used to estimate the curvature term and the rest (factorial vertices) are used to estimate the factorial model. Checking this box will use the adjusted model for the diagnostics, model graphs, point prediction, and optimization.

Output precision (xml/csv): Use to control the output of significant digits to alternate formats.

Default variance ratio: The variance ratio controls the power, precision and optimal build computations for split-plot designs. The default of 1 indicates that the whole-plot variance divided by the subplot variance is 1. Do not change this unless the experiments routinely require a different variance ratio than the default.

Hide build aliases above: By default when a design with aliased three-factor interactions is built, the alias structure is shown before factor information is entered. Changing to main effects or 2FI will show fewer alias checks while changing to 4FI will show more alias checks.

VarTran

See your VarTran documentation for how to set these values. The default settings work best for analyzed data sets saved as VarTran files.

Click the Defaults button to return to the original installed settings for this screen.