Analyzing Variability in DOE
Identify factor settings that produce consistent results
A traditional analysis of a designed experiment helps you determine the factor settings that produce the best average response. You might want to know the settings that produce a target length for an automotive part, or the procedures that result in the shortest resolution times at a call center. But if there is a lot of variability in these length measurements or call times, the best setting on average may not produce the best results.
To ensure your process produces goods or services that meet your customers' expectations, you need to identify the factor settings that not only perform well on average, but perform the most consistently. You can do this in Minitab with the Analyze Variability command.
Why analyze variability
Analyzing average responses in your designed experiment may not give you all the information you need to determine the best factor settings. Here are some examples of how variability can also impact your choice.
 | Here we see how long it takes for operators to process orders after undergoing different training and using different software applications. The average time for existing and new software is about the same, but the existing software clearly shows more variability. To improve customer satisfaction, the call center manager may decide to exclusively use the new software. |
 | Or consider a newspaper printer. The quality manager knows the Type A printing press and the line speed 5 setting produce the highest print quality on average, but also result in excessive variability to the point that some newspapers aren't legible. Given this information, the printer decides to use settings that produce a slightly lower but more consistent level of quality. |
How to analyze variability in Minitab
Using a traditional DOE, a national baked goods company has identified the baking settings that produce the best tasting brownies on average. However, they are receiving complaints about taste consistency. The company decides to conduct another designed experiment, this time looking at variability. They want to know the time, temperature, and pan type that results in the most consistent and best tasting brownie.
First, calculate standard deviations
The first step is to compute the standard deviation for each level of time, temperature, and pan type using Minitab's preprocess response command. You can analyze the variability of any 2-level factorial experiment that has multiple measurements at each factor level setting. The baked goods company replicates the experiment by running each combination of factor settings five times in random order, taking one taste score after each run.
- Choose Stat > DOE > Factorial > PreProcess Responses for Analyze Variability.
- Choose Compute for replicates in each response column.
- Enter the column that contains the response and the columns in which you want to store the standard deviations and counts and click OK.
|  |
Next, find the significant effects
Now you are ready to analyze the variability in your designed experiment.
- Choose Stat > DOE > Factorial > Analyze Variability.
- In Response, enter the column of standard deviation values.
- Under Estimation method, choose Least squares regression to determine which terms are not significant.
- Click Graphs, then choose Pareto and click OK in each dialog box.
|  |
In this Pareto chart of the initial model, none of the two-way interactions are significant. After rerunning the analysis and removing nonsignificant terms one at a time, the final model includes Time, Temp, and Pan. |  |
Lastly, interpret the results
To interpret your final model, rerun the analysis using Stat > DOE > Factorial > Analyze Variability, but this time choose the Maximum likelihood estimation method to obtain moreprecise estimates for the coefficients. Then examine the ratio effects to assess how much variability changes from the low to the high factor setting.
Notice that Time has the largest ratio effect relative to Temp and Pan. Given this information, the baked goods company decides to use the baking time that results in the most consistent taste and sets Temperature and Pan to maintain a high taste score on average. |  |
Putting Analyze Variability to use
Sometimes determining the factor settings that produce the best average response is not enough to ensure high quality goods and services. Excessive variability in your processes and products can result in a higher defect rate and lost customers and profits. So don't just rely on the settings that produce the best result on average. Use Minitab's Analyze Variability to achieve the outcome that effectively and consistently meets your customers' expectations. For more information on Analyze Variability and Minitab's other DOE features, see Minitab Help.