Home / Statistical Tools / Control Charts / G Chart / Options
Options¶
Want to build one first?
How-To works through a finished chart step by step, from sample data you copy off the page straight into Excel. The data has a point that is out of control, so the result has something to read rather than twenty points in a row.
How-To with GroupBy does the same thing with a GroupBy column, so one run produces a separate chart per group.
This page covers the G Chart dialog. It is not built from the same options kit as the variable charts, so several groups you may expect are simply absent rather than disabled, and each absence is called out below at the point where you would look for it. The pane that opens beside the finished chart has its own page, Task Pane. What each option means statistically is on the Math Details page.
The Dialog¶
The dialog is titled G Chart and has three tabs: Data, Limits & Splits, and Rules & Display. There is no Chart: variant row above them, because this chart has one panel and no variants. The buttons along the bottom are Back, Help, Cancel and Finish.
There is no Capability tab. That is not a hidden tab: with no Capability tab there is no specification limit, no target, no capability index and no histogram anywhere in a G chart's output, and the capability path is switched off in the engine as well as absent from the dialog.
Four groups that every variable chart carries do not exist here either. There is no Transformation group, no Control Limits limit-type radios and no manual limits grid, no Historical Values group of the usual five boxes, and no Estimate Sigma group. What replaces the last two is a single Historical pHat (optional): box, described below.
Data Tab¶
The calculation type sits at the top as Excel and GroupBy, with a Data Types button beside it. Below that are two slot lists, Data Columns (required) and X-Axis Labels (Optional), and under GroupBy a region with Available Items and GroupBy Order.
Three things are specific to this chart:
- Only two data types are accepted in the Data Columns list: Count and DateTime. A Continuous or Nominal column cannot be used. Use the Data Types button to set a column's type if it was read as something else.
- Checking more than one data column produces more than one chart, one per column, each using the same settings. There is no subgroup concept on this chart at all: one row is one event, so there is no subgroup size, no subgroup identifier and no subgroups-across-columns mode.
- The minimum is 3 counts, or 4 dates. Three counts is the engine's floor. Four dates produce three counts, because the first date only establishes the starting point and contributes no gap.
Counts or dates¶
A Count column holds the gaps directly: the first value is the count between the start of the record and the first event, or between the previous event and this one, in whatever unit you counted.
A DateTime column holds the date each event occurred, and Quantum XL does the subtraction. Three details matter:
- The gaps are whole days: each difference is truncated downward, so two events 3.9 days apart contribute 3. If you need the fraction, the T Chart keeps it.
- The first date establishes the first event and produces no point, so N dates give N minus 1 plotted counts.
- The dates must be in chronological order. One out-of-order row stops the analysis with a message naming both row numbers and both dates. It does not sort them for you.
When the input is dates, the Y-axis title changes from Count Between Events to Days Between Events.
Limits & Splits Tab¶
This tab carries three groups: Control Limits, Baseline Estimation and Split Control Limits.
Control Limits¶
| Control | Type | Default | What it does |
|---|---|---|---|
| Number of Standard Deviations (k): | text box | 3 | Selects which quantiles of the geometric distribution the lines are drawn at |
| Historical pHat (optional): | text box | blank | Supplies the event probability outright instead of estimating it from the data |
The sigma multiplier does not become a distance on this chart. There is no standard deviation being multiplied. The number you type is converted into a pair of tail probabilities through the normal distribution, and the lines are the geometric quantiles at those probabilities. At the default 3 that is the 0.135th and 99.865th percentiles. Entering 2 pulls the limits in to the 2.28th and 97.72nd percentiles. The value must be greater than zero.
Historical pHat must be strictly between 0 and 1. It is the equivalent of the historical mean and sigma boxes on other charts, expressed the way this distribution is parameterised: one number fixes every line on the chart, and the data then has no influence on where they sit. Leave it blank and the probability is estimated from the gaps you supplied. A value of 0 or 1, or anything outside that range, is refused.
Baseline Estimation¶
| Control | Type | Default | What it does |
|---|---|---|---|
| All Subgroups | radio button | selected | Estimates from every event |
| First N Subgroups | radio button | not selected | Estimates from the first N events only |
The count is a count of events, not worksheet rows, and it is only used for the estimate: every event is still plotted. The limits from those first N events are then extended across the whole chart.
Supplying Historical pHat makes this group irrelevant, because nothing is being estimated.
The count must be a whole number of 4 or more. Type less and the box refuses it with "The first N subgroups must be a whole number of 4 or more." and returns to its previous value.
Four is one minimum for all four of these dialogs, and it is deliberately stricter than the engines behind them, which stop in different places: the G and T calculations need three included values because each estimates a distribution parameter, while the p, np, u and Laney calculations need two because they only pool counts over sample sizes. On top of that, a G or T chart reading a column of dates and times works on the gaps between the dates, so four dates give three gaps. One minimum of 4 satisfies every one of those, and it means the box no longer accepts a number that Finish would then reject.
Split Control Limits¶
| Control | Type | Default | What it does |
|---|---|---|---|
| None | radio button | selected | One set of limits for the whole chart |
| Column | radio button | not selected | A new phase begins wherever the chosen column's value changes |
| Manual Breaks | radio button | not selected | A new phase begins at row numbers you list |
| Split Column: | drop-down | empty | The column read under Column mode |
| Manual Break Rows (1-based, start of split): | text box + Add / Remove Selected | empty | The list of phase starting rows |
| Split Labels (shown beside each divider): | grid | empty | One editable label per phase |
Each phase gets its own \(\hat{p}\), its own center line and its own limits, all estimated from that phase's events alone. A divider is drawn at every boundary and carries the label you typed. The summary table gains a Per-Phase Breakdown block listing each phase's numbers.
Rules & Display Tab¶
Out-of-Control Rules¶
Only Rule 1 is on when the dialog opens. This is the opposite of the variable charts, where all eight tests are on, and it is worth knowing before you wonder why a run of points on one side was not flagged.
| Control | Default | Threshold box |
|---|---|---|
| Rule 1 - Beyond K sigma from center | on | none |
| Rule 2 - N consecutive points on same side of center | off | N = 9 |
| Rule 3 - N consecutive increasing or decreasing | off | N = 6 |
| Rule 4 - N consecutive alternating up/down | off | N = 14 |
| Rule 9 - N consecutive zero counts (Benneyan) | off | N = 4 |
Two of those rows deserve a note.
Rule 1 on this chart means "outside the drawn limits", and since the lower limit is 0 at any realistic event rate, it can only fire upward: a gap longer than the model expects. A gap shorter than expected cannot be flagged by Rule 1, because there is nowhere below 0 to be.
Rule 9 exists only on the G chart. A zero count means two events with no opportunity between them, back to back. Four zeros in a row, at the default threshold, is the Benneyan test firing. The other charts in Quantum XL have no equivalent, and the T chart does not show this row at all.
Rules 5 through 8 are not on this dialog. They read the zone lines, this chart has none, and they are disabled in the engine for it regardless of anything the caller passes.
Every threshold box accepts a whole number of 2 or more. A smaller value is refused with a message box naming the rule, for example "The rule 2 number of points must be a whole number of 2 or more.", and the box goes back to the value it held before. Earlier versions of the G chart refused in silence.
The chart does not tell you which rule fired
Every rule you tick is tested, and the results are combined into a single verdict per point: this point broke at least one enabled rule, or it did not. Points that broke one are drawn in a different color from the rest, and that is the whole of the report. There is no per-rule label, no per-rule marker, no per-rule column, and no list of violations anywhere in the chart or the worksheet. A point that breaks four rules at once looks exactly like a point that breaks one, and the output never names either.
So when you turn several rules on, read a marked point as an instruction to go and look at the chart, not as a diagnosis. If you need to know which pattern caused a particular signal, the practical method is to switch the other rules off and rebuild the chart.
Display Options¶
| Control | Type | Default | What it does |
|---|---|---|---|
| Reference Lines: | grid with Value, Label, Color, Thickness, plus Add Reference Line and Apply | empty | Draws a horizontal line at a value you choose |
There is no zone line option on this dialog, and this chart draws no zone lines. The distribution behind the limits is skewed, so a line a third of the way to the limit is not a standard deviation and the two sides are not the same distance from the center; four lines labelled 1 sigma and 2 sigma would say something untrue. The four values are still computed, and the construction is on the Math Details page, but nothing is plotted and no out-of-control test on this chart reads them.
A reference line is drawn in the chart's own units, which are counts.
Task Pane¶
The Control Chart pane that opens beside a finished chart has its own page: Task Pane.
Changing a column's Data Type can take it out of a slot, and the dialog says so. The column lists on this tab hold only the columns their slot accepts. Change a column's type in the Data Types list to something the slot no longer accepts and the column leaves that slot, even if it was ticked, and a message appears naming the column, its new type, the slot it left, and what that slot does accept:
The column 'Rejects' is set to Continuous, so it has been removed from Defectives, which accepts Count.
The slots and what each one accepts: Defectives and Sample Size on a p chart take Count; Defects on a u chart takes Count, and Inspection Size takes Continuous, Integer or Count; Data on a G chart takes Count or DateTime.
If the column leaves more than one slot, a second sentence follows: "It has also been removed from slot, which accepts types." The chart is then built without that column in that role, so read the message rather than pressing Finish and wondering where the series went.
When a number box refuses what you typed¶
A refused number box now tells you why. Type something this box cannot use, then leave it, and a message box opens naming the box and what it wanted:
The box name must be what it wanted.
The title is the analysis name, for example "G Chart" or "Type 1 Gage Study". A box that wants a proportion adds a second sentence, "It is a proportion, not a percent, so type 0.05 for 5 percent.", because typing 5 for five percent is the mistake it is there to catch. The box then goes back to the value it held before, so nothing is lost and nothing wrong is carried into the run.
Text that is not a number at all gets the same message. Type a word, a stray letter beside a figure, or nothing but a minus sign into any number box on this dialog, leave the box, and a message box opens reading:
The box name must be a number.
The box then shows the value it held before. Before this, a box given text simply would not take it: the entry was dropped on the way to the setting, so none of the rules below could be reached and nothing was said. The box looked as though it had accepted what was typed until the moment the focus moved.
These boxes on this dialog refuse with a message:
| Box | What it must be |
|---|---|
| Number of Standard Deviations (k): | a number greater than 0 |
| Historical pHat (optional): | greater than 0 and less than 1, and the message adds "It is a proportion, not a percent, so type 0.05 for 5 percent." |
| First N Subgroups count | a whole number of 4 or more |
| N = on rules 2, 3, 4 and 9 | a whole number of 2 or more. The message names the rule: "The rule 9 number of points must be a whole number of 2 or more." |
| Value in the Reference Lines grid | a number of 0 or more. A G chart plots a count of opportunities, so there is nothing below zero to draw a line at. |