Design of Experiments Guide  • The Fit Two Level Screening Platform • Overview of the Fit Two Level Screening Platform
Previous
 • 
Next
Overview of the Fit Two Level Screening Platform
The analysis of screening designs depends on the principle of effect sparsity, where most of the variation in the response is explained by a small number of effects. Under this principle, effects with small estimates are used to estimate the error in the model. This then allows one to test whether the larger effects are active.
You can analyze data from a screening experiment using Fit Model (Analyze > Fit Model) or Fit Two Level Screening (Analyze > Specialized Modeling > Specialized DOE Models > Fit Two Level Screening). Use the following guidelines to select the appropriate modeling platform:
•
If your factors are all two-level and orthogonal, all of the statistics in the Fit Two Level Screening platform are appropriate.
•
If you have data from a highly supersaturated main effect design, the Fit Two Level Screening platform is effective in selecting active factors, but it is not effective at estimating the error or the significance. The Monte Carlo simulation to produce p-values uses assumptions that are not valid for this case.
•
If you have a categorical or a discrete numeric factor with more than two levels, the Fit Two Level Screening platform is not appropriate.JMP treats the associated model terms as continuous. For such factor, the variation is scattered across main effects and polynomial effects. In this situation, it is recommended that you use the Fit Model platform.
•
If your data are not orthogonal, the constructed estimates in the Fit Two Level Screening Platform are different from standard regression estimates. JMP can identify large effects, but it does not effectively test each effect. This is because effects are artificially orthogonalized as they are entered into the model, making effects that enter early in the model appear less significant than in standard regression See Order of Effect Entry.
•
For mixture designs, the Fit Two Level Screening platform is not appropriate. See instead Fitting Mixture Designs in Mixture Designs.

Previous
 • 
Next
Help created on 7/12/2018