What is the benefit of "bagging" in a random forest? Get "more" data from our dataset -> less "variance" in random forest