Statalist


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

RE: st: confidence intervals for ratio of predictions-- bootstrap vs. parametric methods?


From   "Daniel Waxman" <dan@amplecat.com>
To   statalist@hsphsun2.harvard.edu
Subject   RE: st: confidence intervals for ratio of predictions-- bootstrap vs. parametric methods?
Date   Fri, 12 Oct 2007 21:31:49 -0400

For the record, I realized that I was bootstrapping the wrong thing.
Here is (a minimally simplified version) of what I meant to do ...

. boot rr_univ=(invlogit($mb_notindicated_predictors)/invlogit($mbneg_predictors)),
/*
      */  reps($reps) saving(multiboot_notneg_$i,replace): /*
      */   logistic outcome zlog zero mbpos int_zlog_pos

. estat bootstrap

(no more red 'x's)

So the question was ... how to explain that these CIs
appear so much better than those generated as follows?

. logistic outcome zlog zero mbpos int_zlog_pos
. predictnl rr=invlogit($mbpos_predictors)/invlogit($mbneg_predictors),se(se_rr)
. gen ub = rr + 1.96*se_rr
. gen lb = rr - 1.96*se

(and is it reasonable to assume that with a whole lot of reps, the
bias-corrected bootstrapped CIs are in fact better?)


Where:

. global mbpos_terms _b[_cons] + _b[int_zlog_pos]*`constant'  + /*
*/ _b[int_zero_pos] + _b[zlog]*`constant'  + _b[zero] + _b[mbpos]

. global mbneg_terms _b[_cons] + _b[zlog]*`constant' + _b[zero]


Perhaps the question is more clear now (?)


Daniel
*
*   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