Process Analysis

Why SPC?

See also Variation and Process Adjustments

Is it reasonable to expect a process to be statistically in-control from the very beginning?

See Statistical Control by Monitoring and Adjustment (Box) for insights.

  • "If you have a house, you know that you must work hard to keep it habitable - the tiles on the roof, the paint on the walls, the washing machine [...] all need attention from time to time. A car, a friendship, and our own bodies must similarly be continually nurtured or they will not remain in shape very long. The same is true for industrial processes. If left to themselves, machines do not stay adjusted, components wear out, and managers and operators forget, mis-communicate, and change jobs. Thus a stable stationary state is an unnatural one and its approximate achievement requires a hard and continuous fight. Both process monitoring and process adjustment can help achieve this. Both are likely to be needed." (Box, page 1)

Very few processes start out in control, or will stay in control, without effort.

Other potentially useful resources (may need screening and review prior to use):

Relationship between Statistical Process Control (SPC) and Time Series.

  • The most common SPC methods (X-bar charts, p-charts, and similar) assume that the results of each subgroup sample are statistically independent of each other. The alternative is data that is autocorrelated. In this case, common SPC methods are not usually appropriate.

  • For one example, see the article attached below "ASQ-using-time-series-on-the-production-floor.pdf".

  • Another example is given in the article attached below "ASQ Why Statisticians Model Data (includes example of high freq SPC and autocorrelation).pdf".

  • See Statistical Quality Control (Montgomery) chapter on "SPC with autocorrelated process data".

  • See Statistical Control by Monitoring and Adjustment (Box).

Which control chart to use? See article attached below "ASQ Which Control Chart Should You Use.pdf", also available from ASQ.