Use Analyze Variability to analyze the standard deviations of repeat or replicate response measurements to detect dispersion effects in your data. This enables you to identify factor settings that produce less variable results. You can also use Analyze Variability to store weights to use when examining the location effects of your response in Analyze Factorial Design. Note that when Minitab fits a linear model with the data, it transforms the standard deviations using the natural log.

You can analyze a 2-level factorial design that includes either repeat or replicate measurements of your response. Minitab provides two methods for fitting your model: Least squares regression and maximum likelihood estimation. You can also generate effects plots to help you determine which factors are important and residual plots to assess model adequacy and assumptions.

A building material manufacturer is developing a new insulation product for construction. The manufacturer studied four factors that were believed to impact the strength (Strength) of the insulation (see StatGuide for 2-level factorial designs). The four factors are:

·    material type (Material)

·    injection pressure (InjPress)

·    injection temperature (InjTemp)

·    cooling temperature (CoolTemp)

While conducting the strength experiment, they decided to collect extra samples to examine the effects of the four factors on the variability in insulation strength. The variation among the samples represents the natural variation in the process when the settings remain the same, plus measurement variation. They collect six repeat measurements of strength at each combination of factor settings and analyze the natural log of the standard deviation of the repeats. In the experiment, they test 16 base runs with 6 repeats, resulting in 96 total samples.

Data: InsulationStrength.MTW (available in the Sample Data folder).