Comments on: Bootstrapping: estimating confidence intervals, standard errors, and bias http://data-sorcery.org/2009/07/04/bootstrapping/ Sat, 05 May 2012 13:46:05 +0000 hourly 1 http://wordpress.com/ By: liebke http://data-sorcery.org/2009/07/04/bootstrapping/#comment-149 Wed, 09 Dec 2009 14:55:16 +0000 http://incanter.wordpress.com/?p=561#comment-149 Yeah, I didn’t provide much detail in terms of the bias estimate.

Bootstrapping is not the source of the bias. Rather, it is used to estimate the bias of the original estimate of the median.

When bootstrapping, we treat the original sample as if it were the population, and then take repeated samples (with replacement) from it and estimate the median of each sample. The resulting collection of medians represents an estimate the sampling distribution of the median of x.

We then take the difference of the mean of this distribution of medians and compare it to the original estimate of the median. The difference of the two is the bias of the original estimate. The mean of the bootstrapped sample is being treated as unbiased estimate of the population median.

Hopefully, that didn’t add to your confusion :)

David

]]> By: Ben http://data-sorcery.org/2009/07/04/bootstrapping/#comment-148 Wed, 09 Dec 2009 10:47:06 +0000 http://incanter.wordpress.com/?p=561#comment-148 What is the source of bias in bootstrapping? I could not follow where this is coming from. Thanks

]]>