Time Series Analysis (CUSUM): Difference between revisions

From Met Dynamics
Jump to navigation Jump to search
(Migrate protected Model Theory content)
 
imported>Scott.Munro
m (Document current Met Dynamics Models behavior (W015))
 
Line 68: Line 68:
* <math>y_t</math> is the original observation (data point) at time <math>t</math>
* <math>y_t</math> is the original observation (data point) at time <math>t</math>
* <math>T</math> is a constant, typically a target value or the mean of historical data
* <math>T</math> is a constant, typically a target value or the mean of historical data
* <math>k</math> is the slope of the V-Mask arm, an optional parameter (V-Mask limits <math>S_{\rm H}</math> and <math>S_{\rm L}</math> are not returned if <math>k</math> is omitted)
* <math>k</math> is the reference value. Omitting it returns only <math>C_t</math>; supplying it also returns <math>S_{\rm H}</math> and <math>S_{\rm L}</math>. These two columns are zero when <math>k=0</math>.
* <math>C_t</math> is the value of the cumulative sum (CUSUM) at time <math>t</math>
* <math>C_t</math> is the value of the cumulative sum (CUSUM) at time <math>t</math>
* <math>S_{\rm H}</math> and <math>S_{\rm L}</math> are the high and low V-Mask parameters, respectively, and the <math>^*</math> superscripts indicate they are optionally-returned
* <math>S_{\rm H}</math> and <math>S_{\rm L}</math> are the upper and lower cumulative statistics, respectively; <math>^*</math> indicates optional results. The function does not evaluate an alarm threshold.
   |}  
   |}  
|  
|  

Latest revision as of 03:52, 23 October 2026

Description

This article describes a method for preparing a Cumulative Sum (CUSUM) control chart.[1][2]

Model theory

Nuvola apps important blue.svg.png This content is available to registered users. Please log in to view.

Excel

The CUSUM control chart may be invoked from the Excel formula bar with the following function calls: 

=mdTimeSeries_CUSUM(TimeSeriesData as Range, T as Double, Optional k as Double = 0)

Invoking the function with no arguments will print Help text associated with the model, including a link to this page.

The input parameters and calculation results are defined below in matrix notation, along with an example image showing the selection of the same cells and arrays in the Excel interface:

[math]\displaystyle{ \begin{align} \mathit{TimeSeriesData} & = \begin{bmatrix} y_1\\ \vdots\\ y_t\\ \end{bmatrix}\\ \\ \mathit{T} & = \big [ T \big ]\\ \\ \mathit{k} & = \big [ k \big ]\\ \end{align} }[/math]
[math]\displaystyle{ \begin{align} \mathit{mdTimeSeries\_CUSUM} & = \begin{bmatrix} \begin{bmatrix} C_{1}\\ \vdots\\ C_{t}\\ \end{bmatrix} & \begin{bmatrix} S_{\rm H,1}\\ \vdots\\ S_{\rm H,t}\\ \end{bmatrix}^* & \begin{bmatrix} S_{\rm L,1}\\ \vdots\\ S_{\rm L,t}\\ \end{bmatrix}^* \end{bmatrix} \end{align} }[/math]

where:

  • [math]\displaystyle{ y_t }[/math] is the original observation (data point) at time [math]\displaystyle{ t }[/math]
  • [math]\displaystyle{ T }[/math] is a constant, typically a target value or the mean of historical data
  • [math]\displaystyle{ k }[/math] is the reference value. Omitting it returns only [math]\displaystyle{ C_t }[/math]; supplying it also returns [math]\displaystyle{ S_{\rm H} }[/math] and [math]\displaystyle{ S_{\rm L} }[/math]. These two columns are zero when [math]\displaystyle{ k=0 }[/math].
  • [math]\displaystyle{ C_t }[/math] is the value of the cumulative sum (CUSUM) at time [math]\displaystyle{ t }[/math]
  • [math]\displaystyle{ S_{\rm H} }[/math] and [math]\displaystyle{ S_{\rm L} }[/math] are the upper and lower cumulative statistics, respectively; [math]\displaystyle{ ^* }[/math] indicates optional results. The function does not evaluate an alarm threshold.
Figure 1. Example showing the selection of the TimeSeriesData (blue frame), T (red frame), k (purple frame), and Results (light blue frame) arrays in Excel. The [math]\displaystyle{ S_{\rm L} }[/math] and [math]\displaystyle{ S_{\rm L} }[/math] parameters are returned in this example as k is specified.

References

  1. ↑ NIST/SEMATECH, 2012. e-Handbook of Statistical Methods, http://www.itl.nist.gov/div898/handbook/, accessed 24 March 2022.
  2. ↑ Napier-Munn, T.J., 2014. Cunning solutions to process improvement. 12th Mill Operators' Conference, Townsville, Qld, 1-3 September 2014.
Retrieved from "https://wiki.metdynamics.com.au/index.php?title=Time_Series_Analysis_(CUSUM)&oldid=18667"