Related Topics:

Augmenting Designs

Changing Design Settings

Design Evaluation

The design folio control panel includes a tool that allows you to evaluate your design and compare it to other designs. By evaluating a design before you implement it, you can avoid using time or resources on an inefficient experiment. For example, if the tool shows that a design is unlikely to detect any main effects, you can increase the numbers of replicates and re-evaluate before performing the experiment. If you have already implemented a design, you can enter the response data and calculate it before performing the evaluation, which will produce more accurate results.

Specifically, the design evaluation tool allows you to:

IMPORTANT: This tool assumes that the response values obtained from the experiment will follow a normal distribution. Thus, it is not applicable to R-DOE, which is intended for life data that is not normally distributed.

For an example of how you can use this feature to compare two designs, see Design Evaluation Example.

Selecting the Settings for the Evaluation

The Design Evaluation tool is available on both the Design tab and the Data tab.

Use the Evaluation Settings page of the control panel to configure the evaluation. The page will differ slightly depending on which tab you access it from, as shown next.

Follow the steps below to define the evaluation settings.

Note that the detection power is calculated in terms of the effect per standard deviation (i.e., the amount by which the effect exceeds the standard deviation, whatever it is), so the standard deviation itself does not affect the calculation when you solve for power. However, in this case, the total effect (Effect per Std Dev * Standard Deviation) will be shown in the Power Study table in the detailed results.

Performing the Evaluation and Interpreting the Results

After you select the desired settings, go to the Main page of the control panel and click the Calculate icon in the Design Evaluation area.

The value you selected to solve for (i.e., power or effect) will appear in the area as shown above. In this example, the probability of detecting the given amount of effect is 79.96% for main effects and 51.64% for 2-way interactions. Note that when there are multiple effects, this value refers to the effect that is the least likely to be detected. (For example, if there are two main effects A and B, and the probability of detecting A is 50% and for B it's 60%, then the Main Effects row will show 0.5000.)

To see more details concerning the evaluation (including the powers calculated for individual effects of interest), click the Detailed Results link at the bottom of the area to view the Evaluation Results window.

 

© 1992-2015. ReliaSoft Corporation. ALL RIGHTS RESERVED.