Each tree is grown by taking a bootstrap sample of N objects chosen at random with replacement from a training set containing N objects, so that the same object may appear more than once in the sample.
Each classification tree is generated by selecting a bootstrap sample of the data, and at each split predictor variables are randomly selected.
This may be generalised by drawing a bootstrap sample of the observed data and re-fitting the imputation model p (y | x ; θ ) to the bootstrap sample, yielding a parameter θ ∗ and hence imputations y ∗ drawn from p (y | x ; θ ∗ ) (Royston, 2004, 2005).
Random Forest (RF) [ 31, 32] is a classification algorithm using sets of random decision trees which are generated by a bootstrap sampling for decision and voting.
A single bootstrap analysis can be performed by 1) generating a bootstrap sample, denoted X1, … X n, by randomly reordering the original values X1, …, X n, 2) calculating the bootstrap CUSUM based on the bootstrap sample, denoted S0, …, S n, 3) calculating the S diff, and 4) determining, whether the bootstrap difference S diff is less than the original difference S diff.
A Random Forest is a collection of classification trees that are randomized by training on a bootstrap sample of the training data and also using only a subset of M (< N) of the variables.
In each bootstrap instance, we created a bootstrap sample by selecting 142 observations with replacement from the original data, which on average are composed of roughly 90 unique patients.
We then created a bootstrap sample by randomly sampling the blocks in the dataset, and computed the bootstrap replicates of the relevant summary statistics of the expression levels.
We obtained a new replication of original dataset (a bootstrap sample) by 1000 random draws of individual subject's data (with replacement) from the original dataset and we fitted the final model to each new dataset.
