A one-sample test of means
We finally have enough knowledge under our belts to perform a null-hypothesis significance test using the bootstrap. In fact, given what we already learned, it can scarcely be easier!
As a note, I prefer to use the the bootstrap mainly as a method of generating confidence intervals and illustrating uncertainty in population parameter estimates, and not as a tool for NHST. But, at least as a demonstration, we'll see a few examples of it being used for hypothesis testing here.
For ease of comparison, let's repeat the one sample test that we performed in Chapter 6, Testing Hypotheses. Recall, that the precip built-in dataset contained the precipitation (in inches) of a sample of US cities. We wanted to know if the mean of the population US precipitation was significantly discrepant from the precipitation average of the rest of the world – a value that we, quite unjustifiably, and arbitrarily, said to be 38 inches.
The one sample t-test was performed thusly:
> t.test...
