To review, in this chapter we explored another powerful method for interval estimation and, to a certain extent, hypothesis testing. First, we put this technique in context to be appealing to the similarity with the resampling simulations we performed in earlier chapters. We learned that the bootstrap is a lot like building sampling distributions from population data, but by sampling with replacement from our sample data instead.
We saw examples of the results from the bootstrap procedure, and noted that it is often congruent with results from parametric interval estimation techniques. In spite of this, we learned about what makes the bootstrap different from other alternatives, what makes it special, and an honest look at some of its drawbacks.
After performing the bootstrap manually, to get a thorough handle on how the procedure works, we learned how to perform it in a more elegant and extensible fashion using the boot
package. We saw that the objects returned from the boot
function...