Re: st: Person-year files, survival analysis, and sample weights

[Steven Samuels <sjsamuels@gmail.com>]

You should use the probability weights ( -stset- with the [pweight=])   
option) before all the analyses that you mention.
From the Manual, it appears that -strate- duplicates the  
functionality of -stptime- and offers more options. Certainly you  
should check your -stptime- results with -strate-.

I'm not sure what you mean by a "person-year" file. If is just an - 
stset- file -stsplit- by single years, then size alone could make it  
undesirable. Also, I can't think of a continuous-time analysis where  
it would be needed.

Steve

Steven J. Samuels
sjsamuels@gmail.com
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783


On Dec 30, 2010, at 10:00 AM, Raj Kapoor wrote:


Hello,

I am using weighted survey data to estimate unadjusted hazards of
death using stptime. My sample comprises of a single baseline wave
that is linked with mortality follow-up data. For various purposes, I
created a person-year file using stsplit. Therefore, my data structure
comprises multiple episodes/records per subject. One reason I have
done this is that I would like to estimate age-specific death rates (I
created a time-varying variable for age).

My specific question is whether it is appropriate to use the survey
sample weights—w/o any adjustment—given that I now have multiple
records per individual?

Similar concerns often arise when estimating various continuous or
discrete time models after creating a person-year (person-time)
file--how should one deal with the sample weights?
So, any general guidance on this topic will also be greatly appreciated.

Thanks in advance!

R

*
*   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/