Hi Everyone,
I am currently doing a comparison with GLM and GBM using R. Of course, I can do a performance index as point comparison from both analysis. However I need an equivalent index of R-square (from GLM) in GBM.
Why can't you compute R-square for GBM. As I understand, the dependent variable is a scale variable (let me know if I am wrong) and the method can compute the predictions, which is all you need to compute R-square.
Hi Ales. Thanks for that. However What I wanted is the R-square from the model itself rather than the R-square between the predicted and observed.
Just wondering, why would you want that, especially if comparing different models. Those "model based" pseudo R-Square could be incompatible.
Exactly the point you raised, however my professor is asking me to find that R-square equivalence in GBM. I already researched about it and was told to use the MSE. However it is still difficult to compare, so I needed someone to guide me or direct me in finding a way just for comparing purposes. Like you mentioned, the pseudo-R square can be misleading that is why I need to be careful
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R