An R Companion to Applied Regression by John Fox & Sanford Weisberg

Author:John Fox & Sanford Weisberg
Language: eng
Format: epub
Published: 2018-09-28T09:07:45.687485+00:00

Unlike confint (), Confint () reports the estimated coefficients along with the confidence limits. In this problem, the confidence intervals produced by confint (), which wrongly assume constant error variance, are unrealistically narrow in comparison to the intervals produced by Confint () using a robust estimator of the coefficient covariance matrix.

5.2.2 Bootstrap Confidence Intervals

One of the primary applications of the bootstrap is to confidence intervals. The bootstrap is attractive in this context because it requires fewer distributional assumptions than other methods. If b1 is an object produced by the Boot () function, then setting the argument vcov.=vcov (b1) in a call to Confint () uses the sample covariance matrix of the bootstrap replicates to estimate coefficient standard errors, otherwise computing Wald confidence intervals in the standard manner. In particular, for linear models, quantiles of the t-distribution are used for confidence intervals when the quantiles of the normal distribution are more appropriate. The intervals produced are therefore somewhat too wide, an issue we consider unimportant because the justification of bootstrap confidence intervals is in any event asymptotic.

For the Transact data, we obtain bootstrap confidence intervals that are very similar to those produced in the preceding section with a robust estimate of the coefficient covariance matrix. Reusing the bootstrap samples in betahat.boot that we computed in Section 5.1.3, we find

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.