Generally you sample with the same size as your number of observations, you
just sample with replacement. Your standard deviations should decrease as
you increase the sub-sample size, because your sample size is getting
larger. We sample with replacement so we can simulate multiple draws from
the population while still maintaining our actual sample size.
Brandon
2009/3/29 Lee, Clarence <clee at hbs.edu>
Hi all,
When you?re using bootstrapping to obtain the confidence intervals and the
significance of your regression coefficients, does the size of subsample
that you sample from in each iteration matter? I noticed that the standard
deviations of my coefficients decreases as the size of the subsample
approach my full sample. What?s the rule of thumb on this? In problem set
1, we only used bootstrapping to obtain the mean and standard deviations of
the coefficients, but not the confidence intervals
Thanks guys,
Clarence
_______________________________________________
gov2001-l mailing list
gov2001-l at
lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l