R2X different with missing values and OPLS/O2PLS (Q682)
The information in this article applies to:
- SIMCA-P+ 12
- SIMCA-P+ 12.0.1
Symptoms
For an OPLS/O2PLS model built on a dataset with missing values, the R2X value displayed in a list or plot from the Plot/List menu differs from the one displayed in the model window.
Cause
R2X can be calculated in two different ways:
A: R2X = 1-ss(X-TP’)/ss(X)
B: R2X = ss(TP’)/ss(X)
For datasets with full X matrices (no missing values) A and B are equivalent. When there exist missing values however, the residuals are not orthogonal to the model which means that neither A nor B is mathematically correct.
Expression A is used in the model window and B in Plot/List.
Workaround
Use the R2X-value found in the model window.
Status
In future versions of SIMCA only expression A will to be used resulting in a more consistent R2X value for models with missing values.
Date modified: jun 28, 2010
Article type: Problem
Comments?
If you would like to comment on this article, please
use this form.
