How to use the I-MR chart maker
- Enter the data as one measurement per row, oldest first — paste from a spreadsheet, type, or upload a .csv / .xlsx file. An optional first column of labels (batch number, date) is detected automatically.
- Choose where the limits come from: all points, the first k points (a Phase I baseline extended over later data) or known values μ₀ and σ₀. Points with an assignable cause can be excluded from the calculation while staying on the chart.
- Choose the tests (Western Electric or Nelson) and read the verdict, both charts, the list of signals and the warnings. Add report details and print or save a PDF.
When to use an I-MR chart
Use the individuals and moving range chart for continuous data when a subgroup would make no sense: one reading per batch, tank or shift; slow processes; destructive or expensive tests; automated 100% measurement where consecutive parts are almost identical; and business data such as daily yield or weekly scrap cost. If you can sensibly take several parts at once, an X̄-R or X̄-S chart is more sensitive to small shifts and far less dependent on normality. For counts of defectives or defects use the attribute charts.
Formulas and constants
The moving range is the absolute difference between consecutive values, MRᵢ = |xᵢ − xᵢ₋₁|. This tool always uses a span of 2, the standard choice, so N points give N − 1 moving ranges. X̄ is the average of the individual values and MR̄ the average moving range.
| Chart | Centre line | UCL | LCL |
|---|---|---|---|
| Individuals (I) | X̄ | X̄ + 2.66·MR̄ | X̄ − 2.66·MR̄ |
| Moving range (MR) | MR̄ | 3.267·MR̄ | 0 |
The constants are those for a subgroup of n = 2:
| Constant | Value | Meaning |
|---|---|---|
| d₂ | 1.128 | σ̂ = MR̄ ÷ 1.128 |
| E₂ = 3 ÷ d₂ | 2.66 | 3σ̂ expressed as a multiple of MR̄ |
| D₄ | 3.267 | Upper limit of the moving range chart |
| D₃ | 0 | No lower limit for ranges of two |
| d₄ | 0.954 | σ̂ = median MR ÷ 0.954 (median option) |
Sigma is deliberately estimated from the moving ranges, not from the overall standard deviation s. The moving range captures short-term, point-to-point variation, so a shift or drift in the data does not inflate the limits and hide itself. The tool shows both values; s clearly larger than σ̂ is itself a sign of an unstable mean.
Median moving range option
One or two huge moving ranges (a spike) inflate MR̄ and widen the limits. As an alternative you can base the limits on the median moving range M̃R: σ̂ = M̃R ÷ 0.954, individuals limits X̄ ± 3.145·M̃R and moving range UCL = 3.865·M̃R. It resists outliers but is less efficient on stable data, so the average moving range remains the default.
Known standards
With known μ₀ and σ₀ the individuals limits are μ₀ ± 3σ₀; the moving range chart has centre line 1.128·σ₀ and UCL 3.686·σ₀.
Out-of-control tests
All selected tests run on the individuals chart. A signal is reported at the point that completes the pattern, and zone tests are evaluated separately on each side of the centre line with σ = (UCL − CL) ÷ 3.
- Beyond the limits (WE 1, Nelson 1): one point above the UCL or below the LCL.
- 2 of 3 beyond 2σ (WE 2, Nelson 5): the point is beyond 2σ and so is at least one of the two before it, same side.
- 4 of 5 beyond 1σ (WE 3, Nelson 6): the point is beyond 1σ and so are at least three of the four before it, same side.
- Run (WE 4: 8 points; Nelson 2: 9 points) on one side of the centre line; a point exactly on the line breaks the run.
- Nelson 3: six points in a row steadily rising or falling. Nelson 4: fourteen points alternating up and down. Nelson 7: fifteen points in a row within 1σ. Nelson 8: eight points in a row more than 1σ from the centre line on either side.
The moving range chart is tested with rule 1 only. Each value appears in two consecutive moving ranges, so the moving ranges are correlated with each other, and their distribution is strongly skewed; run and zone tests on that chart would fire far too often.
How to read the charts — and two warnings
A point outside the limits on the individuals chart, or a run on one side, indicates a change in the process level. A point above the UCL on the moving range chart marks an abrupt jump between two consecutive values; a single spike usually produces two large moving ranges in a row. Investigate and document an assignable cause before excluding any points from limit estimation; keep the original observations visible.
Non-normal data. Averages of sufficiently large, independent subgroups may be closer to normal; this is not guaranteed for small subgroups or strongly skewed data. Individual values do not benefit from averaging. With skewed data (flatness, runout, cycle times, contamination) an individuals chart produces false alarms on the long-tail side. The tool runs an Anderson–Darling test and warns when normality is rejected; confirm with the normality test or a histogram, and consider a transformation.
Autocorrelation. The limits assume consecutive values are independent. Readings taken close together from a tank, an oven or a tool that wears are often positively autocorrelated; the moving range then understates sigma, the limits become too tight and the chart signals constantly. The tool reports the lag-1 autocorrelation r₁ and warns when it is beyond ±2/√N.
Worked example (the sample data)
Thirty batches of a coating are measured for viscosity. The mean is X̄ = 83.42 and the 29 moving ranges average MR̄ = 1.631.
- σ̂ = 1.631 ÷ 1.128 = 1.446
- Individuals chart: UCL = 83.42 + 2.66 × 1.631 = 87.76; LCL = 83.42 − 2.66 × 1.631 = 79.08
- Moving range chart: UCL = 3.267 × 1.631 = 5.33
Batch 12 (88.2) and batch 30 (88.0) are above the UCL — rule 1 signals. The largest moving range, 5.1, stays below 5.33.
The formulas follow Montgomery’s Introduction to Statistical Quality Control and the AIAG SPC reference manual; the tool reproduces Montgomery’s mortgage-loan processing cost example (X̄ = 300.5, MR̄ = 7.79, limits 279.78 and 321.22, moving range UCL 25.45). Once the chart is stable, assess the process against its specification with the process capability calculator.
Frequently asked questions
What is an I-MR chart used for?
An I-MR chart monitors a process when only one measurement is available at a time, such as one reading per batch, per shift or per day. The individuals chart tracks the process level and the moving range chart tracks short-term variation between consecutive readings.
Where does the 2.66 in the I chart formula come from?
It is 3 divided by d2 for a subgroup of two, 3 / 1.128 = 2.66. The average moving range divided by 1.128 estimates the process standard deviation, so the mean plus or minus 2.66 times the average moving range is the mean plus or minus three sigma.
Why not use the ordinary standard deviation for the limits?
The overall standard deviation includes any shifts, trends and outliers in the data, so limits based on it widen exactly when the process is unstable and can hide the problem. The moving range measures point-to-point variation and is much less affected by changes in the process level.
How many data points do I need for an I-MR chart?
At least 20 to 30 individual values are generally recommended before limits are treated as reliable, and more is better because individuals charts are based on little information per point. Limits from fewer points can be used as trial limits and should be recalculated later.
Does my data have to be normally distributed?
The individuals chart is much more sensitive to non-normality than an X-bar chart because there is no averaging. Moderate departures are tolerable, but strongly skewed data cause false alarms. Check normality first, and transform the data or use a different chart if the distribution is clearly not normal.
What moving range length does this tool use?
A moving range of span 2, the absolute difference between each value and the one before it. This is the standard choice and the one the constants 2.66 and 3.267 are based on.