Configuring: Cross Correlation

Decide on the field to set as the target field - this is the field to which other input fields will be correlated. Then decide on whether to calculate the lag of these input fields over a time series or over an index. In each case, configure the number of steps and step size, which defines how far back in time the correlation should be calculated.

By trending the resulting dataset, the peaks displayed on the graph shows the points of greatest correlation, and therefore the desired process delay.

Worked example: Cross correlation

Steps to configure

  1. Select target and fields: Specify a target field and the fields for which to calculate the lags.

  2. Configure lag calculation: Specify the type of delay for the fields.

  3. Execute: A new dataset will be created, containing a lag field and the selected fields with the lagged values.

Step 1: Select target and fields

Select the target field of the dataset from the drop down menu of available fields.

  • Select the fields to be lagged from the list of available fields.

  • Selecting a field as the target deselects it from the available fields window.

  • Default: The first alphabetically listed field is selected as the target and no fields are selected for the lag calculation.

  • Minimum configuration: A target and one field to be lagged must be selected.

Step 2: Configure lag calculation

Select to calculate the lag over time or an index. This is the maximum delay to calculate and consider for process delays on the target fields.  Set this value to the maximum known limit for a delay in the process.  

Note: This operation does not support milliseconds. First use the Resample operation to remove milliseconds present in your data source.

  • Specify the number and size of the steps. This is the number of points to be calculated in the specified period.  Setting this value too low might cause the lag estimation calculation to miss or skip the process delay.  Setting this value too high might cause the lag estimation calculation to take very long.

Time lag:

  • This will calculate the lag over a time period.

  • This value is always specified in seconds regardless of the sampling period in your data.

  • A timestamp field should be selected if the dataset contains more than one timestamp column.

  • Calculating lag over a time span takes longer to execute than calculations over an index base.

Index:

  • This will calculate the lag over an index range, such as the row number.

  • If no overlap exists between the step number and size and the reference dataset, all results will be bad quality (NULL).

  • Default: Index as base is selected and no configuration is done. The first timestamp field will be selected if you use time as base.

  • Minimum configuration: Step number and size must be > 0.

Step 3: Execute

A new dataset will be created, containing the lag field and the selected fields for lagging - the values of these fields are the cross correlation values. The greatest value indicates the point of greatest correlation, and therefore that corresponding lag field value is optimal, either in terms of time or row number.


Related topics:

  

CSense 2023- Last updated: June 24,2025