Subgroups for X-bar and R charts
See also ARL and ATS
See also Sampling and Randomization
Notes about selecting subgroup specimens at one point in time, or selecting specimens across the width of the subgroup interval.
Two approaches to taking subgroup specimens from a process [see p. 290, section 8.2 of Statistical Quality Control (Grant)]
(1) Take all specimens at the beginning of each subgroup interval.
This method decreases the chance for variation within a subgroup and increases the chance for variation between subgroups.
Better suited to process analysis and control.
(2) Take specimens throughout the subgroup interval.
This method increases the chance for variation within a subgroup and decreases the chance for variation between subgroups.
The R chart may show out of control points due to a shift in the process mean during a subgroup interval.
Better suited to decisions regarding acceptance of product.
The first method can miss the occurrence of compensating shifts that might happen between subgroups. Grant suggests the question "whether or not two compensating shifts are likely to occur between subgroups."
Add notes about appropriate stratification. (e.g., multi-cavity molds; lanes across the width of a film web)
Article from Minitab: http://www.minitab.com/uploadedFiles/Shared_Resources/Documents/Articles/being_in_control.pdf