Locked lesson.

Upgrade

About this lesson

This lesson introduces the concept of weighting the data being tracked in a control chart based upon other recent data values. The advantages and disadvantages of this approach are discussed as the weighting principles are introduced.

Quick reference

Time Weighted Control Charts

There are several types of time weighted SPC charts that modify the current data value with historic data.  These are used with variable data parameters to detect small shifts in the mean.

When to use

These charts are often used after a process has first been brought under statistical control by using a different chart.  These are then employed to detect small shifts in the mean, especially if the process has high levels of common cause variation creating noise in the system.

Instructions

Time weighted control charts are created using the control chart process discussed in an earlier module.  The data on these charts is modified from the actual subgroup data to include as aspect of historical data.  These control charts were recent additions to the family of statistical control plots and they do not follow the same pattern as the earlier Shewhart control charts.  The X axis is still sequential time or subgroups.  The Y axis is the modified time-weighted value.  These charts are normally centered vertically on the mean value (or target value).   They still include upper and lower control limits. 

While there are many time weighted control charts, the two most commonly found in business are the CUSUM and the EWMA.  Both will detect small shifts to the mean better than the traditional Shewhart control charts.

  • CUSUM is better at detecting long term drift in the mean.
  • EWMA is better at detecting near term drift in the mean.

Because of the weighting, these charts are less susceptible to noise in the system.  That is why they are an excellent choice for monitoring the mean with a system that has high common cause variation.  These charts will tend to suppress the impact of that noise.

The calculation needed to create the data set and the control limits for these charts is more complex than with the Shewhart control charts.  While it is difficult to do these manually, most statistical software applications have the formulas and can do the calculation and create the control chart plots.

Hints & tips

  • These are a good follow-on control chart to use once a process is under statistical control.  These will quickly identify small shifts in the mean.  The down side is that they will be slightly slower to detect large shifts.
  • The math on these can get tricky. If you are planning to use one of these charts, I strongly recommend Minitab or another statistical software application.
Login to download
  • 00:04 >> Hello, I'm Ray Sheen.
  • 00:06 Now, there's a special category of control charts that were developed after
  • 00:10 Shewhart had laid the groundwork for SPC.
  • 00:14 These are called time-weighted charts, and, although,
  • 00:17 there are quite a few variations they follow the two basic categories.
  • 00:21 Let's do a quick overview of these.
  • 00:24 >> So the first thing you're probably asking is,
  • 00:27 what does time-weighting really mean?
  • 00:29 Time-weighted control charts integrate historical data into the data values and
  • 00:33 control limits that are associated with the current values.
  • 00:36 You may be thinking that the control limits already do that.
  • 00:39 Well, these charts do these in some different and much more impactful ways.
  • 00:44 Because of that these charts will be more sensitive to small shifts in the data.
  • 00:48 But nothing comes free, they will often be less sensitive to large shifts.
  • 00:54 These charts are normally used with variable data,
  • 00:56 but there is no accompanying range or standard deviation chart.
  • 01:00 Now, don't worry the control limits will be kinda funky and
  • 01:04 they will provide some of the insight of those other charts.
  • 01:07 Like all of our other control charts the X axis is time or
  • 01:10 sequential units, the Y axis though is different.
  • 01:14 It's either the measured point or the subgroup mean,
  • 01:17 if we're using a subgroup of data.
  • 01:19 But it'll take those values and
  • 01:21 modify them some from what was originally recorded.
  • 01:25 The charts still have control limit lines on them, but
  • 01:28 the mean vet line is not quite the same.
  • 01:30 In fact, instead the charts stay centered around a mean value.
  • 01:35 So let's look, again, at the Control Chart Selection Decision Tree.
  • 01:39 Recall that we start with the decision concerning what type of data it's
  • 01:42 appropriate to monitor for this process.
  • 01:45 Now, here in this case we're going to be using variable data, but it's
  • 01:49 the Time-Weighted data cuz we're trying to check for small shifts in the data.
  • 01:54 There are two types of time-weighting, one is accumulative sum, or CUSUM, chart.
  • 01:58 And the other is exponentially weighted moving average, or EWMA, chart.
  • 02:03 CUSUM's advantage is with long term trends, and
  • 02:06 the EWMA's advantage is with short term trends.
  • 02:10 And, of course, we've already discussed all the other types of control charts in
  • 02:14 previous modules.
  • 02:15 Let's take a few minutes to discuss time-weighting.
  • 02:18 Time-weighting is based upon our understanding of the importance of recent
  • 02:22 data points.
  • 02:24 A data point is viewed not only as a value in its own right, but
  • 02:28 also in the context of what has gone on before.
  • 02:31 Was there a trend underway and does this point continue that trend?
  • 02:35 Another value of time-weighting is it has a tendency to suppress the impact
  • 02:40 of noise in the system.
  • 02:41 Small basic oscillations are damped out of the data and the impact of outliers is
  • 02:45 reduced, in part, due to the Central Limit Theorom, but also in the time-weighting.
  • 02:51 So the effect is that we see trends sooner, however,
  • 02:54 I don't want to paint too rosey a picture.
  • 02:56 Time-weighted charts are more complex to create and read.
  • 03:00 They don't completely follow the pattern of the traditional
  • 03:03 Shewhart control charts that we've discussed before this.
  • 03:06 In fact, where attribute data is concerned,
  • 03:08 I recommend you do any time-weighting at the time of data collection.
  • 03:12 And then use a traditional Shewhart style chart.
  • 03:15 But we will look at these two charts with variable data as we get to the next
  • 03:19 modules.
  • 03:21 >> Time-weighted control charts are newer additions to the SPC family of charts.
  • 03:25 They provide important insight and some special process categories, but
  • 03:30 they are more complex than standard control charts.

Lesson notes are only available for subscribers.

PMI, PMP, CAPM and PMBOK are registered marks of the Project Management Institute, Inc.

Gift this course