Palisade Knowledge Base

HomeTroubleshooting@RISK for Excel: Simulation"The correlation matrix ... has multiple inputs assigned to the same matrix position."

4.56. "The correlation matrix ... has multiple inputs assigned to the same matrix position."

Applies to: @RISK 5.x–7.x

When I run my simulation, I get the following message:

The correlation matrix ... has multiple inputs assigned to the same matrix position. Instances can be used to correlate multiple sets of inputs with the same matrix. Continue the simulation anyway?

What is wrong, and how do I fix it?

Correlation is specified by RiskCorrmat property functions within the @RISK distributions.  If you have two or more groups of inputs to be correlated using the same matrix, then your RiskCorrmat functions need a third argument, called an instance.  We recommend you click No and edit your workbook. Please see this article for more on instances and how to create them: Same Correlation Coefficients for Several Groups of Inputs.

However, if you're really trying to use this form of correlation to create a correlated time series, please see the help topic "Correlation of Time Series" in @RISK 6.0 and above.  (You need the Industrial Edition of @RISK 6.0 and above to use time series.)

What happens if I just click Yes and let the simulation proceed?

Please download the example SameMatrixPosition.xls (below).

Suppose you specified the same 2×2 matrix to correlate A16:A17 to each other and B16:B17 to each other, without specifying instances.  Then @RISK will indeed use that matrix to correlate A16 with A17, and to correlate B16 with B17.  But to accomplish this, since you correlated those two groups using the same matrix without instances, @RISK will actually use the same underlying random numbers for A16 and B16, giving them a correlation coefficient of 1 with each other. And the same is true of A17 and B17. These correlations between the groups are probably not what you want.

In the same example, D16:D17 and E16:E17 are correlated with instances, avoiding unwanted between-group correlations.

Last edited: 2015-10-15

Downloads

This page was: Helpful | Not Helpful