Home / Statistical Tools / Control Charts / Levey-Jennings
Levey-Jennings¶
From Excel click...
QXL Stat Tools Tab > Control Charts > Variables > Charts as Individuals > Levey-Jennings
A Levey-Jennings chart is a single panel that plots each measurement in the order it was collected, one point per observation. The panel is titled Levey-Jennings Chart. There is no second panel: no moving range, no range, no standard deviation.
Two things separate it from every other individuals-form chart in Quantum XL, and both are on this page. Its control limits are built from the overall standard deviation of the data rather than from a short-term within-subgroup estimate, and it is tested with the six Westgard multi-rule tests rather than the eight Shewhart tests.
Levey-Jennings has a dialog of its own rather than being a variant of another chart, so the dialog title is Levey-Jennings Chart and matches the ribbon entry you clicked. The worksheet it writes is named Levey-Jennings, with no "Chart" on the end. That makes it the one control chart whose worksheet tab does not carry the word Chart, while the title cell written at the top of that same worksheet still reads "Levey-Jennings Chart". Every option in the dialog is described on the Options page.
When to use it¶
- Your measurements arrive one at a time: one value per run, per day, per specimen, per batch. There is nothing to subgroup.
- You want the band to describe the total variation in the data. The limits are placed at the mean plus and minus a multiple of the overall standard deviation, computed from every observation about the mean, so they widen when the data drifts rather than staying tied to the short-term spread between neighboring points.
- You want the Westgard rule set: 1-2s, 1-3s, 2-2s, R-4s, 4-1s, and 10x, all six on when the dialog opens.
If you want a moving range panel beside the individuals panel, or limits built from a short-term estimate of the standard deviation, use an I-MR chart or the Individuals chart instead. This chart offers no choice of sigma estimator and no capability analysis.
How it works¶
Every usable row in your measurement column becomes one plotted point, in worksheet order. The center line is the mean of those observations.
The limits sit a fixed number of standard deviations either side of that mean, and the standard deviation used is the overall one: the ordinary sample standard deviation of the individual observations about their mean. That choice is fixed in the calculation engine and is not exposed as an option. There is no Estimate Sigma group anywhere in this dialog, no Rbar, no Sbar, no pooled estimate, no average or median moving range, and no unbiased-constants checkbox. Every other individuals-form chart in the product estimates the standard deviation from how much consecutive points differ; this one does not.
Because one standard deviation serves the whole chart and every point carries the same weight, the limits are straight lines rather than the stepped ones you see on a subgrouped chart.
The chart is then tested with the six Westgard rules. The chart has one panel, so all six run on it. Five of them are tests the other charts also run, under different names and different thresholds; R-4s is offered on this chart alone. It is a single-stream adaptation of the classic two-material range rule: instead of comparing two control materials measured in the same run, it looks at a consecutive pair of points on the one series and fires when the pair straddles the center line with one point beyond the plus 2 sigma line and the other beyond the minus 2 sigma line.
Three settings change what the limits describe rather than how they are computed. Historical Values lets you supply the mean, the overall standard deviation, or both, instead of estimating them. Estimate Limits From computes them from the first few points only and extends them across the whole chart. Split Control Limits divides the chart into phases, each with its own center line and limits.
Two things this chart does not offer at all: there is no Transformation control, and the engine refuses a transformation outright if one reaches it; and there is no Capability tab, with the capability index and the histogram forced off in the engine as well as hidden in the dialog.
The exact formulas are on the Math Details page.
Output¶
The analysis writes one worksheet per chart, named Levey-Jennings. Under GroupBy each sheet gains " - {GroupName}". On the sheet, from the top:
- A User Input block in the first two columns, holding the analysis title "Levey-Jennings Chart", the measurement column or columns, the X-label column, the split column and the GroupBy columns where you used them, and whether the calculation type was Excel or GroupBy. Modify Existing Analysis reads it back. Only your data selections are recorded here, not your option settings.
- The Levey-Jennings Chart panel, plotting every observation with its center line and control limits, plus the zone lines, reference lines, and phase dividers if you asked for them. Points that broke an enabled Westgard rule are drawn in a different color.
- A summary table to the right, listing the panel's center line, its upper and lower control limits, the count of out-of-control points, and the count of points you marked as outliers, followed by three rows that are specific to this chart: Mean, Sigma (overall), and CV%, the standard deviation expressed as a percentage of the mean. A split chart adds a per-phase breakdown with the same rows. There is no within-subgroup sigma row, because the chart never forms one.
- No capability block and no histogram, on any Levey-Jennings sheet.
When the analysis finishes, the Control Chart task pane opens on the right. See Options for what it can and cannot change.