AI Glossary / Statistics and probability / Intermediate

Rational subgroups

Definition

Rational subgroups are deliberately formed groups of observations used in statistical process control so that within-subgroup variation represents a meaningful short-term source of variation and between-subgroup comparisons can reveal changes of interest. The grouping is based on knowledge of the process, such as timing, machine or batch conditions. Rational does not mean that any convenient set of several readings will do. The choice of subgroup determines what variation a control chart treats as ordinary within-group variation.

What it means

An average chart and a range or standard-deviation chart answer related but different questions. The subgroup average tracks location, while the within-subgroup spread helps characterize variation over the conditions represented inside that group. If measurements from substantially different process states are mixed together, the within-subgroup spread can be inflated and changes between states obscured. Conversely, an artificially narrow subgroup can fail to represent the variation relevant to the intended monitoring question.

How it works

Define the process characteristic, sampling frequency and type of change to be detected. Form subgroups whose members share appropriate operating conditions, often observations collected close together in time. Keep the grouping rule consistent and record contextual information. Monitor both subgroup location and spread using a chart suited to the design and assumptions. Investigate unusual patterns before estimating stable limits. If the process has several machines, streams or hierarchical sources of variation, consider whether separate charts or a more explicit variance structure is needed.

Three practical examples

Example 1: A manufacturer measures several consecutive pieces at scheduled times to monitor short-term production behavior. Example 2: A laboratory groups repeated measurements under the same immediate conditions to track instrument precision. Example 3: A multi-machine operation considers machine-specific subgroups rather than mixing all machines in each group. The right choice depends on the question. Consecutive sampling is common, but it is not a universal rule that overrides knowledge of cycles, dependence or relevant sources of variation.

Think of it this way

Think of taking a set of snapshots during one stable moment and comparing those sets across time. If photographs from different moments are mixed into every set, real changes become part of the apparent variation within each set. The grouping then answers a different question. Rational subgrouping is the statistical version of choosing which snapshots belong together before drawing conclusions from their differences.

Worked example

Suppose three measurements from an early process state are [9, 10, 11] and three from a later state are [19, 20, 21]. The subgroup means are 10 and 20, and both ranges are 2. The grouping makes a large location change visible relative to the short-term spread. Now mix the same six values into [9, 11, 20] and [10, 19, 21]. Their means become 40/3 and 50/3, approximately 13.333 and 16.667, while both ranges become 11. No measurement changed, but mixing states increased the apparent within-group variation and reduced the difference between group means. Two subgroups are far too few to establish reliable control limits; this calculation illustrates the effect of grouping, not a complete control-chart analysis.

Illustrative case study

Fictional case study: Cedar Packaging combines measurements from two filling heads into each subgroup. Engineers notice that the resulting ranges mainly reflect differences between heads rather than short-term behavior of either one. They review the monitoring objective and establish separate tracking where appropriate, retaining a view of overall output as well. The case is fictional and makes no claim that separate charts always solve a process problem. It shows why subgroup structure should reflect the actual sources of variation.

Common misconceptions and mistakes

Common mistakes include choosing groups solely for spreadsheet convenience, mixing different process streams without thought, and treating control limits as customer specification limits. Stable within-group variation does not prove stable subgroup means. Repeated readings of one item may measure instrument repeatability rather than variation between produced items. Serial dependence can also affect chart behavior. A subgroup size copied from another application is not automatically suitable for the present process or decision.

Business use

Business use: Rational subgroups improve the interpretability of process monitoring when they reflect how work is produced and measured. Define the sampling protocol jointly with process specialists. Record changes in equipment, materials and operators so chart signals can be investigated. Separate the questions of process stability, measurement precision and specification compliance. Review the subgroup design when production routing or sampling practices change rather than carrying old limits forward mechanically.

People also ask

Must a rational subgroup contain consecutive items?

Not always; the grouping must match the monitoring objective and process structure. Q: Can repeated measurements of one item form a subgroup? A: They can for a measurement-precision question, but they do not represent ordinary item-to-item variation. Q: Do rational subgroups guarantee normal data? A: No; distributional and dependence assumptions still require attention.

Check your understanding

Why did the ranges increase from two to eleven in the worked example?

Answer: Measurements from different process locations were mixed within each group.

Can the example establish control limits?

Answer: No. It deliberately uses only two groups to demonstrate grouping effects, which is insufficient evidence for a stable monitoring baseline.

Related terms

• X-bar charts monitor subgroup averages

Learn next

Reorganize a small time-ordered dataset into several plausible subgroup structures and compare their means and ranges. Explain which variation each structure measures. Then study the assumptions of X-bar and spread charts, including the consequences of dependence. Build a sampling protocol before using software to calculate limits, and retain enough stable baseline information to support the chart.

More in Statistics and probability

• Rashomon effect in machine learning - The Rashomon effect in machine learning is the presence of multiple different models that predict the same data similarly well.
• Ratio estimation - Ratio estimation uses a relationship between a study variable and an auxiliary variable to estimate a population quantity.
• Ratio scale - A ratio scale is a measurement scale with meaningful equal intervals and a meaningful zero, so ratios of values can represent ratios of the measured quantity.
• Realized variance - Realized variance is an ex-post measure of variation calculated by summing squared returns over subintervals within a specified period.
• Recall bias - Recall bias is systematic error arising when people inaccurately or incompletely remember and report past events or experiences.
• Rectified Gaussian distribution - In the clipped-normal sense used here, a rectified Gaussian distribution is the distribution of Y = max(0, X) when X is Gaussian.

Sources

Sources: NIST/SEMATECH e-Handbook glossary, rational subgroups: https://www.itl.nist.gov/div898/handbook/glossary.htm ; NIST, Subgroup Analysis: https://www.itl.nist.gov/div898/handbook/pmc/section6/pmc613.htm . These primary institutional sources connect subgroup structure with process monitoring and sources of variation. The regrouping calculation and fictional packaging case are original illustrations.

Previous: Rational quadratic kernel Next: Rationality