The SHEWHART Procedure


Displaying Stratified Process Data

If the data for a Shewhart chart can be classified by factors relevant to the process (for instance, machines or operators), displaying the classification on the chart can facilitate the identification of special or common causes of variation that are related to the factors. Kume (1985) refers to this type of classification as "stratification" and describes various ways to create stratified control charts.

There are important differences between stratification and subgrouping. The data must always be classified into subgroups before a control chart can be produced. Subgrouping affects how control limits are computed from the data as well as the outcome of tests for special causes (see Tests for Special Causes: SHEWHART Procedure). The values of the subgroup-variable specified in the chart statement classify the data into subgroups. In contrast, stratification is optional and involves classification variables other than the subgroup-variable. Displaying stratification influences how the chart is interpreted, but it does not affect control limits or tests for special causes.

This section describes three types of variables that you can specify to create stratified control charts.

  • A symbol-variable stratifies data into levels of a classification variable.

  • The block-variables stratify data into blocks of consecutive observations.

  • A _PHASE_ variable stratifies data into time phases.

You can specify any combination of these three variables. You should be careful, however, because it is possible to generate confusing charts by overusing these methods.

The data for the examples in this section consist of diameter measurements for a part produced on one of three different machines. Three subgroups, each consisting of six parts, are sampled each day, corresponding to three shifts worked each day. The data are provided in the data set Parts, which is created by the following statements:

data Parts;
   length Machine $ 4;
   input Sample Machine $ Day Shift DiamX DiamS;
   DiamN=6;
   datalines;
 1  A386  01  1  4.32  0.39
 2  A386  01  2  4.49  0.35
 3  A386  01  3  4.44  0.44
 4  A386  02  1  4.45  0.17
 5  A386  02  2  4.21  0.53
 6  A386  02  3  4.56  0.26
 7  A386  03  1  4.63  0.39
 8  A386  03  2  4.38  0.47
 9  A386  03  3  4.47  0.40
10  A455  04  1  4.42  0.37
11  A455  04  2  4.45  0.32
12  A455  04  3  4.62  0.36
13  A455  05  1  4.33  0.31
14  A455  05  2  4.29  0.33
15  A455  05  3  4.17  0.25
16  C334  08  1  4.15  0.28
17  C334  08  2  4.21  0.33
18  C334  08  3  4.16  0.19
19  C334  09  1  4.14  0.13
20  C334  09  2  4.11  0.19
21  C334  09  3  4.10  0.27
22  C334  10  1  3.99  0.14
23  C334  10  2  4.24  0.16
24  C334  10  3  4.23  0.14
25  A386  11  1  4.27  0.28
26  A386  11  2  4.70  0.45
27  A386  11  3  4.51  0.45
28  A386  12  1  4.34  0.16
29  A386  12  2  4.38  0.29
30  A386  12  3  4.28  0.24
31  A386  15  1  4.47  0.26
32  A386  15  2  4.31  0.46
33  A386  15  3  4.52  0.33
;