Of coz you can do that in non-parametric bootstrap method!!!!! One advise, you will be better off to do this by comparing the ratio of (-a1/b1) and (-a2/b2). Things to concern, when you do bootstrap sampling, make sure you can sample the matching pair in two sets(in other words, if the first set is {A, B, C, a, b, c}, and second set must be {A, B, C, a, b, c} in the same iteration) as my understanding to your problem.
For your information, there is two method that I know for doing bootstrap in regression. First method, you can sample X and Y directly. Second method, find the fitted value first and sampling residuals.



雷达卡
京公网安备 11010802022788号







