Model Answers to 08 ANOVA in Six Sigma Statistics using Minitab 17, Green Belt Edition.

Analysis 3

Analysis 4

We are told that changes in these five factors could explain 83.69% of changes in the response. A slight drop from the last experiment. We are also given the optimal factor settings to maximise the response.

After setup of the worksheet the next step is for the trials to be conducted so that the response data can be gathered. Luckily we have a sheet already prepared for you.

The data for this section is  13 DOE.xlsx worksheet ‘Power Model’.

  1. Go to the Minitab worksheet where our modeling DOE worksheet was created. Delete everything but the title bar and then transfer the data from the Excel worksheet into Minitab.
  2. Click Assistant<<DOE<<Analyze and Interpret.
  3. Click on the Fit Screening Model box.

4. A confirmation menu box will appear and ask you to confirm the goal of the experiment. Ensure the goal is set to ‘Maximise the response ‘ and then click OK to produce the 5 page Modeling DOE report.

On the Prediction and Optimization Report, in the top centre, we are given the 95% prediction interval we can expect to achieve when conducting runs at the settings given to maximum power. Those settings are shown as the optimal solution.

Previous Page Previous Page

The Summary report indicates that we would be able to detect small effects with this number of experimental runs

Set-Up 4

In the Session Window we find that the Analysis of Variance table only has significant terms. As we have reduced the number of terms the Lack-of-Fit p-Value has appeared. It tells us that the model does fit the data.

We are then shown the settings that we need to set in order to achieve the maximum response using main effects plots. For all the factors this is the hi level of the factor

Later in the are shown the top five alternate solutions.

The Report Card, not shown here, does tell us that Blocks were not significant and there were no other issues with our study. Curvature was not detected so we do not need to add more experimental points in order to explore the model.