Flexible Features to Refine and Extend the Analysis
In situations where a response is not normally distributed, or where the error variance of each observation is not constant, you may wish to apply a response transformation. This allows you to analyze the response using a different scale (e.g., you could use ln(Y) instead of Y). If you aren't sure whether to use a transformation, DOE++'s Select Transformation tool can help you decide and even recommend a transformation for you.
Optimization for Your Desired Outcome
DOE++’s Optimization Folio provides a collection of powerful tools that you can use to explore the factor level combinations that produce response values within the limits you specify. This includes the ability to search for the factor level combination that produces the most desirable output, as well as the ability to display all the combinations that keep the output within the specified limits.
- The Optimal Solution Plot solves for the most desirable factor level combinations. It can take into account multiple responses with different maximum, minimum or target values. To help you find the truly optimal settings, you can specify how much emphasis to place on each of these values, in addition to specifying how important the optimization of each response is.
- The Overlaid Contour Plot varies two factors and holds the others fixed in order to help you visualize all the feasible factor level combinations that are expected to produce response values within the specified limits.
- For more complicated scenarios where you want to view the feasible factor level combinations, the Dynamic Overlaid Contour Plot allows you to vary three factors or more.
In Version 10, you can now choose to hold factors fixed at specified levels during the optimization. We've also made it easier and more intuitive to specify the optimization settings.
Response Prediction for Untested Treatments
DOE++'s Prediction window makes it easy to enter your own combinations of factor settings to predict the resulting response values — complete with confidence metrics — based on the fitted model.
For designs with more than one replicate, DOE++ allows you to determine the variability of the response(s) across runs and to analyze that standard deviation information. This offers valuable insight into the sources of variation within the experimental data.
With the new Simulation Worksheet, you can link an experiment design to a diagram in BlockSim/RENO to obtain simulated response data. The simulated data can then be analyzed in DOE++ in order to investigate the effect of one or more settings on the simulation results. This empowers you to make useful predictions about how different factors influence responses such as cost or availability, without having to invest the time and resources needed to conduct an experiment.