Skip to content

Home / Shared Math Details / Control Chart Math / Laney Overdispersion (Sigma Z)

Laney Overdispersion (Sigma Z)

This page defines the Laney overdispersion factor \(\sigma_Z\), which corrects an attribute chart's control limits when the data are more (or less) variable than the binomial or Poisson model assumes. Quantum XL reports \(\sigma_Z\) as a diagnostic on every attribute chart and applies it to the limits on the Laney P' and Laney U' charts.

Notation

Term Description
\(v_i\) the plotted statistic at subgroup \(i\) (proportion for p, defects-per-unit for u)
\(\text{CL}\) the center line (\(\bar{p}\) or \(\bar{u}\))
\(\sigma_i\) the ordinary binomial or Poisson standard deviation at subgroup \(i\)
\(z_i\) the standardized subgroup value
\(\overline{MR}_z\) the average moving range of the \(z_i\)
\(d_2\) the range-to-sigma constant for a moving-range window of two, \(d_2 = 1.1284\)
\(\sigma_Z\) the overdispersion factor

Standardized values

Each subgroup is standardized by its own ordinary sigma:

\[ z_i = \frac{v_i - \text{CL}}{\sigma_i} \]

If \(\sigma_i \le 0\) the value is set to \(z_i = 0\).

Overdispersion factor

The factor is the average moving range of the \(z_i\) (a window of two, so \(MR_i = \lvert z_i - z_{i-1}\rvert\)) divided by \(d_2\):

\[ \sigma_Z = \frac{\overline{MR}_z}{d_2}, \qquad d_2 = 1.1284 \]

Any moving-range window that spans a missing subgroup is dropped rather than bridged, and \(\overline{MR}_z\) is the plain mean of the surviving ranges. When \(\sigma_Z\) cannot be computed it is taken to be \(1\) (no correction).

Applying the factor

On the Laney charts the ordinary control limit is used with its sigma scaled by \(\sigma_Z\):

\[ \text{UCL} = \text{CL} + k\,\sigma_i\,\sigma_Z, \qquad \text{LCL} = \text{CL} - k\,\sigma_i\,\sigma_Z \]

Diagnostic use

On the ordinary p, np, c, and u charts, \(\sigma_Z\) is computed and reported but not applied to the limits. A value near \(1\) indicates the binomial or Poisson model fits and the ordinary chart is appropriate; a value meaningfully above \(1\) indicates overdispersion, and the Laney P' or Laney U' chart is preferred.

Used by

See Also

References

  1. Laney, D. B. (2002). Improved control charts for attributes. Quality Engineering, 14(4), 531-537.
  2. Montgomery, D. C. (2013). Introduction to Statistical Quality Control, 7th ed. Wiley.