Summary
From this chapter we learned that the bootstrap can be applied to almost any complex problem, but we also learned that the bootstrap must be adapted for each complex problem. For regression analysis this was done by sampling from residuals instead of the whole data matrix. For times series analysis, the modification of the bootstrap was done by splitting the time series in blocks and resampling within blocks.
We also saw that uncertainty and proper variances can be estimated for data including missing values. This has huge advantages whenever multiple imputation cannot be applied for logistic reasons in a company or organization.
The bootstrap was also applied to complex survey samples drawn with complex survey designs. Here we defined the calibrated bootstrap to adequately estimate the variance of a statistic.
Monte Carlo tests served as a very general tool for hypothesis testing. Data scientists can make use of them for any statistical test. We did not use any theoretical knowledge...
