Locked lesson.

Upgrade

About this lesson

Some of the most important decisions with respect to control charting are the decisions about subgroups and samples. These decisions will dictate the type of control chart that should be used. They also will determine the number of data points in a subgroup. The decisions should be made based upon the characteristics of the process that is being control charted.

Exercise files

Download this lesson’s related exercise files.

Subgroups and Samples .docx
204.1 KB Subgroups and Samples - Solution.docx
203.8 KB

Quick reference

Subgroups and Samples

A critical decision associated with Control Chart selection is the size of the subgroup and the sampling approach to be used when selecting data points within the subgroup.  The values entered into the Control Chart are a descriptive statistics of the subgroups dataset of sample points.

When to use

Every Control Chart has rules and guidelines associated with establishing the subgroup size and selecting sample points within the subgroup.

Instructions

The values plotted on the SPC charts that we are concerned with will be values that represent some characteristic of data associated with a subgroup of the process execution.  The determination of the boundaries for a subgroup should be set so that subgroups represent items that were processed in a similar manner and under similar conditions.

Typical subgroup selection consideration are process management parameters such as:

  • A process batch
  • A process time interval such as a day or an hour
  • A process completion of an assembly or system
  • A fixed number of items that represent a lot or production run

Subgroup selection is an intentional decision that assists operators and managers to effectively manage the process.  Too few subgroups will prevent the quick identification of process issues.  Too many subgroups creates extra work without providing additional meaningful information.

Samples are the specific data points that make up the subgroup data set.  These data points, are descriptive statistics about the dataset are what is plotted on the control chart and are used to calculate the control limits. The actual sample selection is based upon the type of subgroup.  Attribute data SPC charts will normally check every item in the subgroup for its defect status.  Variable data SPC charts will normally take a sample of the items within the subgroup definition.  This sample should be selected at random, not with a set pattern. With attribute data, in some cases we are counting defects and in some cases defectives.

  • Defects are an attribute that does not meet the expectation
  • Defectives are a unit that contains one or more defects. 
  • Example: an assembly unit that misses dimensional requirements in three locations has three defects but is only one defective.

The table below provides ground rules and principles to be used when selecting subgroup size and selecting samples

Hints & tips

  • Establish clear subgroup boundaries so that when operators are collecting data they know into which subgroup the sample data point belongs
  • The most common method for setting subgroup boundaries is time based – hourly, shift, day, or week.
  • The next most common method is a production run, such a s batch or a work order for as set number of units.
  • Consider the Selection Ground Rules column of the table when deciding which type of chart is most appropriate for your process.
Login to download
  • 00:04 Hi, I'm Ray Sheen.
  • 00:06 A critical decision with regards to control charting
  • 00:09 is the selection of subgroups and samples.
  • 00:12 Let's take a few minutes and review the principles associated with those.
  • 00:16 I'll start with subgroups.
  • 00:19 SPC charts will plot values associated with subgroups.
  • 00:23 Each data point represents the next subgroup.
  • 00:26 Now, depending upon the type of chart,
  • 00:28 the subgroup parameters being charted will be different.
  • 00:32 It could be the mean or the standard deviation.
  • 00:35 It could be a count of defects or
  • 00:37 percentage of defective items in the subgroup.
  • 00:40 The parameter will dictate what type of chart to use.
  • 00:44 A subgroup is telling us something about the process.
  • 00:47 So items that are in the same subgroup should have been processed
  • 00:51 in the same manner.
  • 00:53 That way the subgroup value is showing how the process operated
  • 00:57 with that group of items.
  • 00:59 This means that the subgroup selection is driven by how the process is managed.
  • 01:04 If the process is a batch process, the subgroup is each batch.
  • 01:09 If the process is a continuous flow process,
  • 01:11 then a subgroup should be the items flowing through the process
  • 01:15 over a time period, such as an hour, a shift, or a day.
  • 01:19 If a process is related to completing a complex system or
  • 01:24 a big assembly, then the subgroup would be each system.
  • 01:28 The key point to emphasize is that this is an intentional decision.
  • 01:32 Don't just pick a subgroup or
  • 01:33 a subgroup size because that's what you've always done.
  • 01:37 Instead, let the process dictate the boundaries for each of the subgroups.
  • 01:42 Now, within the subgroup there are data points, which we refer to as the samples.
  • 01:47 The samples are all the data points that comprise the data set of the subgroup.
  • 01:52 There may be just one data point or there may be many,
  • 01:55 depending upon the type of subgroup and the control chart selected.
  • 01:59 The selection of the sample data points is based upon the type of subgrouping.
  • 02:04 If your subgroup is a batch or
  • 02:06 a time interval in a continuous flow process, the data points or
  • 02:10 sample should be randomly selected from all available data points in the subgroup.
  • 02:15 If the subgroup is a system or assembly, your data points are normally
  • 02:19 all the possible pass-fail points within that system.
  • 02:23 In some cases, the subgroup may be multiple systems or
  • 02:26 units, so that the data points would be all of the pass-fail checks within
  • 02:30 all of the units in the subgroup.
  • 02:33 This leads us to an interesting point with respect to attribute data and
  • 02:36 that is the difference between defects and defectives.
  • 02:40 A defect is a failure to meet expectations on any measured parameter.
  • 02:45 A defective is a unit that often has multiple measured parameters,
  • 02:49 in which at least one of these parameters is classified as a defect.
  • 02:54 So, for example, if I have a unit with ten measured parameters, and
  • 02:58 three of those do not meet the expectations, I have three defects.
  • 03:02 But I only have one defective unit since all three occurred in the same unit.
  • 03:08 Let's wrap up this lesson with a summary of the key ground rules or
  • 03:12 principles with respect to subgroups and samples for each of the control charts.
  • 03:17 The I-MR chart is the easiest.
  • 03:19 The subgroup size is one, so the sample size is one.
  • 03:22 That means that the parameter is measured every time.
  • 03:25 This is normally used with a process that operates infrequently.
  • 03:30 The XbarR chart has a subgroup for
  • 03:32 a batch or a managed process interval such as a set time period.
  • 03:37 It randomly selects two to ten data point samples within that subgroup boundary.
  • 03:44 The XbarS chart is just like the Xbar chart except that it is costing
  • 03:47 more than ten data point samples within that subgroup boundary.
  • 03:51 The C chart uses a fixed subgroup size that is selected so that when
  • 03:56 every parameter is checked, the average count of defects is greater than two.
  • 04:01 This is often used with the process that has many opportunities for
  • 04:06 a defect, but few actual occurrences.
  • 04:09 The U chart is showing us defective units.
  • 04:12 So the subgroup size of the number of units is selected so
  • 04:16 that when every parameter is checked, there are at least five defects,
  • 04:21 not five defectives, but five defects, in each subgroup.
  • 04:25 The number of units within the subgroup can vary and
  • 04:27 is often based upon a process management parameter, such as units in a day.
  • 04:33 This is particularly good when the flow in the process leads to an erratic
  • 04:37 number of units in the subgroup.
  • 04:39 The NP chart is focused on defectives, not defects, and
  • 04:43 uses a fixed number of units in the subgroup.
  • 04:46 All parameters are checked on all the units in the subgroup.
  • 04:49 So the number of units that comprise a subgroup is selected in such a way
  • 04:53 that the average number of defective units within the subgroup is greater than five,
  • 04:58 and the average number of successful units is also greater than five.
  • 05:02 That may mean you need a very large subgroup in order to meet both of those
  • 05:06 ground rules.
  • 05:07 This is a great control chart to track first pass yield for assemblies or
  • 05:12 systems.
  • 05:14 The P chart is also focused on defectives, not defects.
  • 05:17 And it also has a subgroup size so when all parameters are checked,
  • 05:21 the average number of units with the error rate is greater than five and
  • 05:25 the average number of units of success is greater than five.
  • 05:29 In the P chart, the number of units in each subgroup can vary, so
  • 05:33 it's often used to track error rates in processes.
  • 05:37 The cumulative sum chart has a subgroup size of one and
  • 05:40 therefore a sample rate of one.
  • 05:42 This chart is normally used when the process management needs immediate
  • 05:47 notice of even small shifts in the mean value for the parameter being measured.
  • 05:53 And finally, the exponentially weighted moving average chart, pose the subgroup or
  • 05:58 sampling principles as the I-MR, XbarR, or XbarS chart.
  • 06:03 This chart is often used when a process has a lot of noise in the measured
  • 06:07 parameter that can cause false alarms with traditional control charts.
  • 06:12 Ground rules for sample and subgroup size are critical parameters
  • 06:16 that must be followed when creating control charts.
  • 06:20 Consider your process and the parameter being measured.
  • 06:23 Then pick the control, finally size
  • 06:28 the subgroup and sample accordingly

Lesson notes are only available for subscribers.

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

Gift this course