I would like to do bootstrap aggregation of a model (currently fit with glm()) so that: 1) Data observations are replicated as N bootstrap samples. 2) The specified model is fit to each sample. 3) Error is calculated on out of bag samples. 4) Have an easy way of making model predictions. 5) + all other sensible features that you can inherit from a random forest implementation. I can find plenty of packages (ipred, randomForest, etc) that bag trees but none that will give the bootstrap aggregation features to a general model. For a glm, the closest is probably the package randomGLM, however this does not seem to let you fix the covariates in the glm model. Does anyone know of neat/elegant implementations of the general bagging procedure, or should I write something myself, like this (horrible code): http://stackoverflow.com/questions/21785699/bagging-logistic-regression-in-r Thanks, Alex