
More on the Bootstrap
Trevor Hastie and Robert Tibshirani, professors of statistics at Stanford, continue their treatment of the bootstrap as a resampling method for estimating the variability of a statistic. This segment, part of the Statistical Learning course offered through Stanford Online and edX, extends the basic bootstrap idea introduced earlier, covering practical subtleties such as how the method behaves when the quantity being estimated depends on more complex procedures, and how bootstrap estimates compare with other measures of uncertainty. The presentation uses worked examples and plots to show how resampling with replacement from a dataset can approximate a sampling distribution without relying on strong parametric assumptions. Aimed at students already following the course's earlier material on resampling methods, it assumes familiarity with basic statistical inference and cross-validation concepts covered in prior sessions.