Hello, Regarding imbalanced data: When using sampsize correction to balanced inbalanced data in Random Forests, what are the implications of the algorithm no longer using a bootstrapped sample? For instance, if I set sampsize to 25, 25 for a binary response, in a dataset of N=800, how does randomforest create a tree? Are there any reasons to not set sample size equal? Regarding partial dependence plots: I have gotten conflicting information on how these are created. I believed that these were created for each predictor with all other predictors held at their mean value. In Cutler et al. 2007 (ecology), however, they say that all combinations of the predictor values are used. Thank you in advance, Valerie -- Valerie Steen PhD Student, Ecology, Colorado State University U.S. Geological Survey, Fort Collins Science Center 320-309-7237 [[alternative HTML version deleted]]