ECON 710 Midterm: ECON 710 UW Madison Midterm 2002a
Document Summary
Sample answers: the non-parametric bootstrap estimates the distribution of wi = (yi, xi) by the empirical distri- bution function fn(w). Since the data wi is iid, the edf is consistent and the non-parametric bootstrap is valid. Heteroskedasticity, e. g. random e e2 i | xi is irrelevant. One way to see this is that the edf is drawing the pairs (yi, xi) jointly, so the full conditional distribution of yi given xi matches the data. The only important cavaet is to observe that heteroskedasticity a ects the validity of standard errors, and this is important for the percentile-t bootstrap. The presence of heteroskedasticity means that the standard errors should be calculated using the white formula. Doing so, the t-ratio is asymptotically standard normal (and hence pivotal) which is necessary to achieve an asymptotic re nement: we have. Combined with p and the cmt, we nd.