Correlation matrix affected by missing values (Q826)

The information in this article applies to: 
SIMCA 13.0.2
SIMCA 13
SIMCA 14
Summary: 

When calculating the averages of the variables for the correlation calculations, all values in the workset are used. Only the values that are non-missing in both variables should be included. With a lot of unmatching missing values, the correlations may be inaccurate.

Status: 

Fixed in the next release.

More Information: 

When calculating the averages of the variables for the correlation calculations, all values are used. When calculating the correlations, only the values that are non-missing in both variables should be used.  

Date modified: Sep 21, 2015
Article type: 
Problem