beta coefficients and Prediction


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

Report Cancel

I find that PLS prediction using calibration model was the same (or very similar) as computation (prediction) using beta coefficients obtained from the same calibration model only for mean-centering modeling (or other regression models as well) only when without any preprocessing.

However, when I added some other preprocessing such as averaging or derivatives. the above two results were not the same. Quite different for derivatives.

It appears that, somehow, computed beta coefficients does not compensate for the effect of other tasks averaging or derivatives except regression (PLSR) itself.


About the Author

Thread Reply

  1. External Admin
    19/07/2019 at 10:09 am

    Please briefly explain why you feel this answer should be reported .

    Report Cancel

    When using the regression coefficients for prediction, the same pre-treatments used in the calibration stage must be applied to the data being predicted.

    When doing the prediction directly in Unscrambler (Tasks->Predict->Regression) there is a ‘Pretreatment’ button that can be used to specify which auto-pretreatments (of those saved with the selected model) to apply on the data before doing the prediction. When predicting for original (raw data), the pretreatment boxes should be ticked on, while for preprocessed data, the boxes should be ticked off.

Leave an answer