Discriminant analysis on spectra without PCA?


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

Report Cancel


is it possible to use a LDA directly on spectroscopic data without a PCA for data reduction? Does it make sense?

In general there is always the combination PCA-LDA on spectroscopic data. In literature it is not usual just using the LDA directly. But are there any reasons besides the data reduction step not using the LDA like this?

Best wishes,


About the Author

Thread Reply

  1. 27/01/2020 at 7:49 am

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

    Report Cancel

    LDA does not work well for correlated data, such as spectroscopic data. Using PCA-LDA takes care of this in addition to the issue of having more variables than samples. This is because instead of the original variables, PCA scores, which are linearly independent (uncorrelated),  are the input for LDA.

    An alternative to using PCA-LDA on spectroscopic data could be to select the apices of the peaks as input for LDA instead of the entire spectrum. It might be that the data is still correlated though.

Leave an answer