Create Definitive Screening Design

In many applications, the number of factors that potentially affect process quality is too great to study all factors in detail. The usual goal of a screening design is to identify the most important factors that affect process quality. After screening experiments, you usually do optimization experiments that provide more detail on the relationships among the most important factors and the response variables.

For example, process engineers at a company that sells dairy products design an experiment to study various factors that affect ice cream texture. The experiment includes 9 factors including fat content, pasteurization, temperature, homogenization process, mixing speed, draw temperature, emulsifier, stabilizer, and cooling speed. The engineers create a screening design so that they can determine which factors are most important. Then, the engineers plan to use other designs to study these factors in greater detail.

Perform the analysis

Complete the following steps to specify the design.

Enter the name of your response variable: The worksheet includes a column with this name where you enter the data from the experiment.
Number of factors: In Number of factors, select how many factors to study. The table of factors changes.
Table of factors: Under Name, enter a descriptive name for each factor.; Specify whether each factor is continuous or categorical. For any continuous factors, enter numbers. Enter the lower number for the study in the Low column. For example, to study the temperatures 30 and 40, enter 30 in the Low column and 40 in the High column.; For any categorical factors, enter labels for the levels. Labels can be numbers or text. If you have a text factor and the levels have no natural order, you can specify the levels in any order.
Number of replicates: Select the number of replicates. Replicates are multiple experimental runs with the same factor settings (levels). One replicate is equivalent to the base design, where you conduct each run once. With two replicates, you perform each run twice (in random order), and so on.; Adding replicates can help increase the precision of your model and increase the power to detect effects. To determine how many replicates to include in your design, consider the available resources and the purpose of your design. For example, in a screening design or in sequential experimentation you could begin with the base design (1 replicate) and then consider whether to add replicates after you analyze the data. You can add replicates to your design later with Stat > DOE > Modify Design. For more information on replicates, go to Replicates and repeats in designed experiments.

Example

A quality engineer wants to study the influence that seven input variables (factors) have on the shrinkage of a plastic fastener of a toy. The engineer plans a pilot study to screen these seven factors to determine which factors have the greatest influence on the response. The engineer is primarily interested in which factors are most important for shrinkage.

Choose Stat > DOE > Quick Designs.
Select Select a Screening Design.
Select Create an experiment with 7-48 factors. Select OK.
In the new dialog, in Enter the name of your response variable, enter Shrinkage.
In Number of factors, select 7.

Complete the table with the following settings:

Name	Type	Low	High
Cooling time	Continuous	10	20
Injection pressure	Continuous	150000	250000
Injection speed	Continuous	5	10
Injection temperature	Continuous	180	360
Packing pressure	Continuous	150000	250000
Holding pressure	Continuous	150000	250000
Thickness	Continuous	1.4	1.8

In Number of replicates, select None. Select OK.

The design summary table shows that the design has 17 base runs, which include 1 center point. The worksheet contains the 17 runs in run order, which is random.