[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: what's wrong witt bootstrap in seemingly unrelated regression

From (Jeff Pitblado, StataCorp LP)
Subject   Re: st: what's wrong witt bootstrap in seemingly unrelated regression
Date   Sat, 07 Mar 2009 17:55:24 -0600

Jingjing Li <> is using -bootstrap- with -sureg, isure- but
is getting the 'last estimates not found' error:
> I am doing seemingly unrelated regression, using "sureg()()(),  
> const()isure". Since the number of my observations is very small, I am  
> tring to do it again with bootsrap method. Now I have two questions:
> 1. The first command I used was " bootstrap  _b, reps(500):  
> sureg()()(), const()isure". But it took such a long time, so I changed  
> reps(500) to reps(2) to have a try. At the end there is an error:  
> r(301), last estimations not found.
> Then I dropped "isure", so the command becomes " bootstrap  _b,  
> reps(500): sureg()()(), const()". This time the results came out  
> quickly. Does that mean "bootsrtap" can't be used  with "isure"? Why?  
> Then how can I get the bootsrap results base on my origianl sureg with  
> "isure"? Since the coef. based on sureg with and without "isure" are  
> totally different.
> 2. I find the coef. after bootstraping are exactly the same as the  
> original regression. I am wondering if the only thing I should care  
> about is p value. If bootsrap p value is similar to the original one,  
> it means the original regression results based on small number of  
> observations is correct. Is it right?

Martin Weiss answered Jingjing's second point above, so I'll address the
first point.

-sure- appears unable to replay it's own results when the -isure- option is
specified.  We will fix this and have it ready in a future (probably the very
next) ado-file update.

*   For searches and help try:

© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index