Hello everybody,
I would like to perform a Fisher's Linear Discriminant Analysis using a stepwise procedure in R. I tried the "MASS", "klaR" and "caret" package and even if the "klaR" package (stepclass function) works correctly I would need also the Fisher's value and the global significance level (p-value).
What I would like to do is to replicate this method: http://www.uta.edu/faculty/sawasthi/Statistics/stdiscan.html
Thanks
Maurizio, The reason that you can't find it is that stepwise is very unstable in the sense that small changes in the data make big changes in the variables selected. Lasso or boosting will give much more stable results. See the attached book for details. Best wishes, David
Maurizio, The reason that you can't find it is that stepwise is very unstable in the sense that small changes in the data make big changes in the variables selected. Lasso or boosting will give much more stable results. See the attached book for details. Best wishes, David
Thanks for the pdf and for the reply. I will look at it soon!
Regards, maurizio
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R