Learn more in our free online course:

**Statistical Thinking for Industrial Problem Solving**

In this video, we show how to generate a screening design using the Custom Designer in JMP. We’ll create an optimal screening design for the Heck reaction scenario, with five factors.

First, we select Custom Design from the DOE menu.

We change the response to Yield.

There are five continuous 2-level factors. The factors, and the levels for each factor, are stored in a file, Heck Reaction Factors.jmp. Instead of entering the factors and factor levels individually, we click the red triangle and select Load Factors.

When we do this, the factors and the factor levels are loaded into the Factors outline.

Our next step is to specify the model that we want to estimate.

Suppose that we want to estimate all 2-way interactions. To do this, we select Interactions, 2^{nd} in the Model outline. All 2-way interactions are automatically added to the model.

You can see that all of the 2-way interactions have been added to the model.

The minimum number of runs required is 16, and the default is 20.

What if we don’t need to estimate all the 2-way interactions. For example, we might know, based on prior data or subject matter knowledge, that Sodium Acetate can’t interact with any of the other factors.

We can click these interactions and then click Remove Term.

As we remove each interaction, the number of runs updates. After removing all four interactions with Sodium Acetate, the minimum number of runs required is 12, and the default is 16.

Let’s look at a few other options that are available.

When you design an experiment, you might have some constraints. For example, you might not be able to run certain combinations of factor settings. You can specify these constraints using the options under Define Factor Constraints.

We won’t do this here.

There are many options under the red triangle. Let’s take a look at the Optimality Criterion.

The default optimality criterion for our design is Recommended. When JMP generates this design, it generates a D-Optimal design. D-Optimal designs are often used for screening experiments, where the goal is to identify and precisely estimate important effects.

For information about the different optimality criteria, search for Optimality Criteria at jmp.com/help.

Let’s go ahead and create this design.

To do this, we click Make Design.

You can see the design that has been generated. Because we only have main effects and 2-way interactions, each of the factors is at two levels. Note that there are many possible designs that are equally good, so if you design this experiment, your design might be slightly different.

There are many options under Design Evaluation for evaluating this design.

We’ll go ahead and click Make Table to create this design table.

You can see that this is a 16-run D-Optimal design.

With this design, you can estimate five main effects and six 2-way .