DOE Fractional Factorial Design (Legacy)

When to use this tool

Use the fractional factorial DOE to analyze the effects of multiple factors, each at two levels, on a numerical response. These designs are typically used when you have a large number of inputs and you want to screen out the trivial effects from the significant few that drive the desired output.

The advantage of fractional factorial designs is that they use a subset (fraction) of the full set of possible design runs to estimate the effects, so they are very efficient designs. The disadvantage is that they confound some effects in the process, so not all effects can be estimated, but it is usually possible to select a fraction that gives the best information on the effects of interest while sacrificing the unimportant effects.

Using EngineRoom

Use the Design Wizard to create the experimental design. Once the design (the design matrix) is created in the data sources panel, enter the response variable data into the last column in the matrix. Then, use the DOE Analyzer tool to analyze the design. Follow the example below to learn how to do this in EngineRoom.

Example:

Create and analyze a fractional factorial design to improve customer satisfaction scores with a website design. There are five factors under study, each at two levels:

• Color Format: One/Two
• Marketing Style: Pop-up/List
• Layout: Short/Long
• Live Help: No/Yes
• Testimonials: No/Yes

Steps:

1. Create Design:

• Select Analyze > Design of Experiments > click on the Design Wizard.
• Select the Expert mode.
• For Select the design, click on 2 level Factorial
• Click on the plus '+' button below the two rows already displayed, until you have 5 rows. Then enter the information on the five factors and click Continue:

• In the Setup step, select the Res V, 16 run design and leave the remaining items at their default settings, then click Create Design:

2. Enter Data:

• The design matrix opens up from the data sources panel. It contains all the information about the design, with a blank column at the end; normally you would run the experiment by following the Randomized run order, and enter the response data into this column. However, we already have the response data, so we can simply enter or copy-paste it here, after making sure the run order of both designs is identical (i.e., run #1 in the design we just created is the same as run #1 in the design table provided.) The response data provided in the data file above are sorted in Standard order, so sort the newly created design in the Standard order as well by clicking on the Up arrow in the Standard order column:

• Copy-paste the response (Score) data into the blank column and click the Save Changes button:

3. Analyze Design:

• Now let's analyze the design - select Analyze > Design of Experiments > DOE Analyzer .
• Drag the Design matrix variable from the data sources panel onto the Design matrix drop zone on the Analyzer, and the Score variable onto the Response Variable drop zone.
• The output contains several parts. Click on the 'study setup' button on top right to see the effect plots: Half normal effect plot showing only the main effects B, A and C are significant:

Pareto effect plot showing the same result as above:

• Click on study setup to close the window and select the Summary tab to see the design summary, factor information, Half normal effects in the model and the Alias structure showing which effects are aliased/confounded:

• Click on the Model Output tab - this is the model output along with the residual plots:

• Click on the Factorial Plots tab to see the charts of main effects and interactions. The main effect plots are shown here; click on the interactions and cube plots buttons at the bottom to see the other charts:

DOE Fractional Factorial Design Video Tutorial