CUSUM control chart
Accumulates deviations from a target so that small, persistent shifts in a process mean show up much sooner than on an X-bar or individuals chart (tabular CUSUM: upper C+ and lower C−, reference value k and decision interval h in sigma units).
Menu: Graphs → More charts → Control chart: CUSUM…
When should I use it?
- Detecting small drifts (about 1 sigma) in a process mean.
What are the assumptions?
- Values in time order; independent; sigma estimated from a stable period (or given).
How do I run it in StatLab?
- Choose the measurement and, if measured in subgroups, the subgroup variable.
- Keep k = 0.5 and h = 5 (good for 1-sigma shifts) unless you have a reason to change them.
How do I read the output?
| Output | Meaning |
|---|---|
| C+ / C− | Running sums of deviations above / below the target (C− drawn below zero). |
| H | Decision interval; a sum beyond it signals a shift (red squares). |
Worked example
On the practice survey dataset (File → Sample dataset), this command — which the dialog's Paste button also produces — runs the procedure:
PROC vizCusum /y=score /x='' /k=0.5 /h=5.
Output tables: CUSUM Signals.
Computed by StatLab's validated procedure code when this page was built. The example only illustrates the steps; choose variables that fit your own research question.
What are common mistakes?
- Using CUSUM to judge single large outliers — use a Shewhart chart for those.
- Setting the target from the very data that contain the shift.