[R-sig-eco] NA in predict.randomForest

[Mats Lindegart]

HI!
I am having problems with predictions in a randomForest classification model. Some predictions turn out to NA:s (both classes and probabilities). The strange thing is that when I use predict.all=TRUE I can see that the model produce seemingly valid predictions for all trees (also for those lines that say NA in my predictions). Anyone have an idea of what is going on?

Best regards,
Mats Lindegart


Ass. Prof. Mats Lindegarth
Department of Marine Ecology - Tjärnö
S- 452 96 Strömstad, SWEDEN