The sixpack function lets us generate six charts to give the control charts and capability in a single page. This forms a useful overview of the stability of our process and how well we fit to customer specifications. It also helps to avoid the common error of trying to fit specifications to control charts.
The example we will be using looks at the fill volumes of syringes. We check the capability of the fill volumes against a target fill volume of 15 ml and specifications of 14.25 ml and 15.75 ml.
Within the worksheet, 40 results are collected per day, taken at the rate of five samples per hour. The data is presented as these subgroups across rows; each row representing the results of the five samples within that hour.
The following steps will generate Xbar-R charts, normality tests, capability histograms, and Cpk and Ppk, all on a six-panel graph page:
Volume3.mtw
by using Open Worksheet from the File menu.C2
to C6
into the section under the section Subgroups across rows of.14.25
.15.75
.15
in the Target field.The Capability Sixpack option produces a page of six charts. These give an overview of the process from control charts, capability histograms, and a normality test on the data. This creates a very visual summary page.
The control chart displayed for the sixpack will be chosen from I-MR, Xbar-R, or Xbar-S charts, based on the subgroup size.
Results that can be used are either listed in one column or where the data is laid out in the worksheet with subgroups across rows. Each row has the five samples measured every hour. The drop-down selection at the top of the dialog box is used to specify the layout of the data. See the Capability analysis for normally distributed data recipe for data used in one column.
There are four sets of tools that we can use on capability studies. They are as follows:
Using a normal distribution to fit to nonnormal data will give us an incorrect estimate of capability. Depending on the nature of the results, the direction they are skewed in can cause the capability to be over or underestimated. One method of finding a more accurate estimate of capability is to apply a transform to return the data to a normal distribution.
Transformation of data should only be used if we understand why the data is not normally distributed and we are confident that the natural shape of the data is not normal. See the examples on Box-Cox or Johnson transformation. Transformations are only found in the capability tools for normal distributions.
The Capability Sixpack can also be run on nonnormal data and as a between within study as well. These are not covered here, but the instructions are similar to those here and the capability analysis for data that does not fit a normal distribution.
18.218.224.226