General Statistical FAQ

You can find the basic information on these sampling plans in Duncan's book onQuality Control and Industrial Statistics - publisher is Irwin - or in Grant and Leavenworth's Statistical Quality Control - publisher is McGraw Hill.

Problem:

Our Company uses SPC Direct to perform real-time SPC on the shop floor. A few of the variables that we monitor are either naturally cyclic or are adjusted by process controllers, therefore the corresponding control charts indicate a severe lack of control. Does this mean that traditional SPC charts can’t be applied in such cases?

Solution:

Since the process controller’s job is to constantly adjust one or more variables in order to maintain a desired state of output , it is not surprising that the control charts in question indicate a lack of process stability. In such cases, traditional control charts aren’t useful since we already should be used to monitor the controller and search unusual events.

To modify your existing chart(s), run the Database Manager and select the Auto Regressive Time Series Module. After identifying the station and Part you wish to study, the resulting screen will indicate the variables that show evidence of auto regressive behavior. SPC Direct will assist you in selecting the appropriate ‘model’ for the controller(s) and automatically import the necessary information to the data entry monitor. The resulting control chart(s) will now monitor the residuals between the measured and predicted values, thus looking for ‘unusual’ events.

1. The abbreviation CapSD stands for the "capability standard deviation".

2. The capability standard deviation is an indication of what the standard of the process would be IF the process were to be in the state of statistical control. If the process is not-in-control, then the CapSD will usually be significantly smaller than the actual standard deviation of the data.

3. Custom/QC arrives at this estimate in several different ways:

From the histogram analysis module it uses two different approaches

• The Mean Square Successive Difference estimator of the standard deviation
• the average of the standard deviations of carious long runs of data

From control chart analysis it uses standard deviation estimators such as Rbar/d2 or sbar/c4 or sigmabar/c2

4. These estimates are all trying to give you an estimate of the potential value of the standard deviation of your process - based on the type of analysis you are currently using. They usually are not identical since they are calculated in different fashions. The smallest of the various capability standard deviations is an indication of the amount of variation reduction you may be able to accomplish by controlling the process (in the sense of achieving statistical control).

Problem:

Our company uses Custom/QC for Windows^TM to generate off-line control charts, histograms, and descriptive statistics. Is there any way we can combine all three output types on a single page for a concise management summary?

Solution:

Custom/QC for Windows^TM allows the user to configure the items that will appear on the graphics output page to create customized summary reports.

To create a customized graphics page, select one out of the 13 control chart types from the Control Charts submenu. The output tab on the resulting dialog box contains an area for the Graphics Page configuration. Select the items that you wish to include on the output summary. For example, if an XBar and Range chart, Histogram, And selected statistics. Only the items that you select will appear on the output page. Please note that the quantity of statistics shown depends on the number of graphics page options that have been selected