About Correlation Maps

Because Isight can support large numbers of data sets, you can use correlation maps to sort the relevance of particular parameters and the relationships between the parameters.

Related Topics
Correlation Map Reference Information

You can use correlation maps to:

  • illustrate how the model works through its data flow,

  • identify areas of redundancy,

  • identify weak points, and

  • identify the rules that govern the performance of the system.

Correlation maps allow you to make decisions to focus the efforts of your design on the parameters that have the most effect on a desired outcome.

The following figure shows an example of a correlation map:



The labeled nodes down the diagonal are the parameters used by the model. The inputs (marked by yellow arrows pointing into the squares) are located in the upper left half, and the outputs (marked by blue arrows pointing out of the squares) are located in the lower right. Connecting pairs of these parameters together are the edges that represent a correlation. If the edge is in the upper right half, there is a positive correlation; the edges in the lower left indicate a negative correlation. The lines are further subdivided, with the strongest half of the currently displayed lines represented as a solid line and the weaker half as a dashed line.

The correlations are ranked from –1 to 1. The sign of the value indicates if the correlations are direct (+) or indirect (–); the value indicates the strength of the correlation (0 is the lowest, +/–1 is the highest). The cutoff value indicates the minimum absolute value of the correlations to display. For example, a cutoff value of 0.5 will display all the correlations greater than 0.5 and less than –0.5. You can change the cutoff value to control which correlations are displayed (see Changing the Cutoff Value).