Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: qreg: # of replications and covergence


From   <J.Fischer@lse.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: qreg: # of replications and covergence
Date   Thu, 27 Apr 2006 18:32:40 +0100

Dear Satalist users


I am estimating a quantile regression for various quantiles with about
6000 individual observations and ca. 30 explanatory variables. Since I
have both aggregate(state) and individual level data combined, I thought
it might be better to have bootstrapped standard errors.

One economterician told me that I should use about 1000 replications,
but, on the other hand, I read that for bootstrapping only standard
errors but not the coefficients in a linear model (?) about 100
replications would suffice.

So my first question is: do I need 1000 replications ?

Second, I observed that the bootstrapping procedure does not converge
(with 1000 rep.). Stata then interrupts the bootstrapping process and
starts from the beginning again.
I have already used the wlsiter option to have more repetitions of the
first stage of the regression (now: 20), when a weighted estimation is
carried out. Increasing the number in  the wlsiter option should solve
convergence problems - but is does not in my case.

How can I make this animal converge ? 
(This problem is a bit mitigated when I use a smaller number of
replications (e.g. 100)  but still tendes to persist)

Thanks a lot

Justina
*
*   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/

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index