Statalist


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

Re: st: Re: limiting observations within a program


From   Michael McCulloch <mm@pinest.org>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Re: limiting observations within a program
Date   Fri, 3 Oct 2008 08:47:24 -0700

Thanks Austin. Your two questions each lead to a different solution:
1. One call to the program successfully returns the expected hazard ratio and confidence interval. However, when I increased the bootstrap replications above 3, then the upper and lower bounds of the confidence interval are inflated to 10^7! What does that suggest is happening to the standard errors?

2. Thanks for pointing out the misplaced ` in the iweight statement for stset. When I switch to [iw=`wt_stab_ALL'], bootstrapping my program returns the error message "varlist required". Setting trace on pointed out that this error message was triggered at the command line for indeplist. When I removed that command, which also required me to remove the [tcm=r(tcm)] portion of the bootstrap command line, my program returns the hazard ratio of 1 (which was expected since I know there is no survival difference between the groups.)

This adds one more question, for which I apologize as I'm sure it's an elementary one: when you implement Cox regression in an analysis of two exposure groups where there are **no** failures in either group, what should happen to the hazard ratio?
Michael



Michael McCulloch <mm@pinest.org>:
Have you tried looking at one call to your program?

  set tr on
  msmcox_stageALL_includeIPTWd

Note that [`iweight=wt_stab_ALL'] does not use the weights while
[iw=`wt_stab_ALL'] does.

What are those weights supposed to do for you?  Shouldn't the weights
be decreasing in the estimated probability of treatment for obs who
get treatment, and increasing in the estimated probability of
treatment for obs who don't get treatment?

On Thu, Oct 2, 2008 at 8:46 PM, Michael McCulloch <mm@pinest.org> wrote:
 Thanks for confirming `X', and also for pointing out the nifty
 abbreviations!
 Now, if I could only figure out why adding the -if- statement to my -stcox-
 command:
        stcox tcm if `wt_stab_ALL'!=.

 or dropping the very with missing weight with:
        drop if `wt_stab_ALL' ==.

 will cause the bootstrapping of my program to give the error message:

 "varlist required
 an error occurred when bootstrap executed msmcox_stageALL_includeIPTWd
 r(100);"

  Michael

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   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