With 22 predictors things can get rather fragile.
On very general grounds, which may not apply
in this particular case, I'd say choose neither
if your results aren't similar and you can't explain why.
Nick
[email protected]
Rijo John
>
> I dont know if I should conclude that it is because of the randomness
> that we expect. For example for one of the variable I am
> getting a std error
> of .1238498 in method 2 but .0903295 in method 3 which is
> causing a shift
> in p value from 0.101 to 0.025. Is that strange? If we are to take a
> decision on the significance of a coefficient based on the
> P-value which
> p-value (among the three we get) should we accept?
>
> :You mean differing other than because of
> :randomness?
> :> I was doing a bootstrapping exercise for the same quantile
> :> regression in
> :> three ways.
> :> 1) bs "qreg y x1 x2 x3....x22" "_b[x1] _b[x2]...._b[x22] _[_const],
> :> reps(20)
> :> 2) sqreg y x1 x2 ...... x22
> :> 3) bsqreg y x1 x2 ......x22
> :>
> :> Why am I getting different bootstrap standard errors in each
> :> one of this
> :> method? The stata help says these methods are same. The
> :> coefficients I am
> :> getting are same though. Can anyone tell me which one of
> this is to be
> :> accepted or most widely used?
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
