Confidence intervals

Now that we have this new, empirically determined sampling distribution (the bootstrap distribution), we can provide interval estimates of population parameters in much the same idea as we can with parametrically determined sampling distributions.

For both the parametric sampling distribution of sample means and the bootstrap distribution of sample means, the range of values from the 2.5% quantile and the 97.5% quantile of the sampling distribution represent the range of values for which 95% of the sample means would lie if repeated samples were taken from the population.

Note

Note that the bootstrap, at least as introduced thus far, is a frequentist technique, and therefore does not use credible intervals like those of the last chapter.

In Chapter 5, Using Data To Reason About The World, we saw that we can get 95% confidence intervals by getting the quantile of the t-distribution of our particular sample size, and using this to multiply with the standard error. With the...