When trying to perform prediction I keep getting a warning that not enough variables are selected, although I have selected the same variable set as my calibration model. What are the requirements for selecting columns for prediction in Unscrambler?

Report
Question
views

Please explain why do you think this question should be reported?

Report Cancel

The most common cause is that one or more autopretreatments require more variables than the actual model. If data have already been pretreated, make sure to unselect the autopretreatment option. If not, make sure to select all the variables required by the first transformation selected for autopretreatment.

In Projection, Prediction and Classification, Unscrambler will attempt to map the variables using relative indexes. The following numbers of variables are allowed:

  • Same number as number of original pretreatment variables.
  • Same number as the difference between max and min pretreatment index.
  • Same number or more than max pretreatment index.

While this logic will handle most scenarios, it is still possible to make mistakes. For instance, if the same number or more than the required number of variables are specified, the pretreatment range relative to the first column will be assumed. If this is not the case, prediction will be allowed but results will be wrong.

To be on the safe side, you can let the matrix with prediction data have the same column dimension as the pretreatment matrix (i.e. column 1 is same in both matrices, irrespective of actual input range).

About the Author

Leave an answer