This guide provides instructions on analyzing screening experiments (e.g., fractional factorial) using the Fit Two Level Screening platform. Though the Fit Model platform can also be used, the Fit Two Level Screening platform produces graphs and analysis results designed for identifying the important effects in a fully saturated experiment. This example is a 20 run 2^5 screening experiment generated from the Custom Design platform. This design allows estimation of all main effects and two-factor interactions.
Specify the Model and Analyze
Most experiments designed in JMP will have Screening and Model scripts saved to the data table. The Screening script launches the Screening analysis platform and automatically fits a saturated model.
- Launch the Fit Two Level Screening platform under DOE > Classical > Two Level Screening or Click on the green triangle next to Screening to run the script if saved to the data table.
- JMP fits a saturated model (here, 19 terms plus the intercept). The Contrasts table and Half Normal Plot identify active factors using Lenth Pseudo Standard Error (PSE).
Notes: In screening experiments, we assume that most effects are inactive and their estimates are essentially random noise. The line in the Half Normal Plot is drawn with a slope equal to the Lenth PSE (an estimate of the residual standard error). Most effects are inactive and fall close to this line. Effects that deviate substantially from this line are labeled as active.
- Select Run Model (at the bottom) to launch the Fit Model platform with only the active effects. Note: You need to include the main effects for any factors that are included in a significant interaction regardless of the p-value for that main effect. Add those terms by selecting them in the table while holding the command key. Results include: Effect Summary, Lack of Fit (if replicated points), Summary of Fit, ANOVA table and more.
Other options – such as residuals or normal plot, profiler and interaction plots – are available under the top red triangle.
Tips:
- Individual and Simultaneous p-Values in the Fit Two Level Screening platform are based on Monte Carlo simulation (and will vary).
- An alternative approach to running the analysis is to use Analyze > Fit Model specifying the model or run the Model script. See the “DOE Fractional Factorial Analysis” guide.
Visit Design of Experiments Guide in JMP Help to learn more.