
Variables
Charts
Xbar, R charts allow the process operator to track his/her process average
(Xbar) and process variation (R, for Range) over time. See the example in the
graphic on the Roles & Benefits page. Typically, the operator will take a subgroup of,
say, 5 consecutive samples. You want to make sure it is likely that there will be only a
minimum of variation within each subgroup, so that variation between subgroups shows up on
the chart. After taking the samples for a subgroup, the operator then calculates the mean
(Xbar) and the range (R) of the subgroup. The subgroup average is plotted for on the top
chart, and the range is plotted on the bottom chart.
After 20 or so subgroups are plotted, a grand average (Xdouble bar) of all of the
subgroup averages is calculated and plotted as a horizontal line on the top chart. Also,
an average (Rbar) of all of the subgroup ranges is calculated and plotted on the bottom
chart. The Rbar value can also be used to calculate the Upper and Lower Control Limits
for both charts. These represent the normal limits (+ or  minus 3 standard deviations, or
99.7%) of the population of subgroups.
Any plot points beyond the control limits suggest a special cause of variation. In a
process under statistical control, there should be about as many subgroup averages above
the grand average line as there are below the line. About 2/3rds of the points should lie
fairly close to the grand average line  within the region 1/3rd of the distance to each
control limit. There should not be too many points in a row above or below the grand
average line, or too many in a row increasing or decreasing. (These rules are more
specifically defined in most SPC courses and texts).
Attributes Charts
See a
typical attributes chart (an np chart). Although it is usually preferable to use variables
data and charting, attributes charts can be beneficial when you are dealing with go/nogo
data, qualitative data, etc. This type of charting does not have R (subgroup range)
calculations or plotting. If you are interested in plotting the number of defects per unit
(which usually has many defects), then use a c (with constant sample size) or u (with
varying sample sizes) chart. If it is more appropriate to chart the actual number (np) or
percentage (p) of defective units in a subgroup, then use a p or np chart. The np chart
requires a constant sample size, but it can vary with the p chart. The c and np charts are
the easiest to use; with the varying sample sizes for the u and p charts, you have to
constantly recalculate the control limits.
