R-squared NA for PLSR prediction


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

Report Cancel

I am using UNSCRAMBLER X for soil spectroscopy. I was predicting some values with PLSR, but when I looked at the predicted vs. reference statistics, it showed me “NA” for the R-squared value. I was trying to find out why it doesn’t show zero or the actual value, but didn’t find any answer for it. Is there any specific threshold value below which the R-squared doesn’t get shown anymore, as the correlation coefficient was still shown, so the R-squared should be possible to calculate from that.



About the Author

Thread Reply

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

    Report Cancel

    The R-square statistic in Predicted vs. Reference is the value for the Explained Variance for the selected number of components (the same values you see in the Explained Variance plot in the Regression Overview window when building a PLS model). When this value is negative, which when observed for the validation explained variance is a typical sign of a poor model, the R-square for the selected number of components will be shown as a missing value, NA, in the Predicted vs. Reference statistics.

    For the correlation between predicted Y and Reference Y, refer to the ‘Correlation’ and/or ‘R2(Pearson)’ statistics.

Leave an answer