Statalist


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

Re: st: Survival analysis and weights - what changed in Stata 10?


From   Bryan Sayer <bsayer@chrr.osu.edu>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Survival analysis and weights - what changed in Stata 10?
Date   Fri, 03 Apr 2009 14:11:23 -0400

I haven't looked at what is available in Stata, however I know that they are doable. SUDAAN has it.

Bryan Sayer
Health Data Systems and Analysis (HDSA) Center
Monday to Friday, 8:30 to 5:00
Phone: (614) 442-7369
FAX:  (614) 442-7329
BSayer@chrr.osu.edu
www.s-3.com



Sergiy Radyakin wrote:
as a follow up, I just found what changed here:
http://www.stata.com/help.cgi?whatsnew

 13.  sts list does not support standard errors and confidence intervals
         when the data are stset with pweights.  However, when option at()
         was also specified, standard errors and confidence intervals were
         reported, but these were inappropriate for sampling-weighted data.
         As such, they are no longer reported.

Is there any way to get the correct SEs then?

Many thanks,
   Sergiy Radyakin


On Fri, Apr 3, 2009 at 12:56 PM, Sergiy Radyakin <serjradyakin@gmail.com> wrote:
Dear All,

I am doing survival analysis based on survey data (with stratification
and pweights). I am using standard commands -stset- and -sts list- to
obtain the estimates of the survivor function and corresponding
standard errors.
If I include weights in the stset command in Stata 10, I don't get the
standard errors of the survival probabilities.
I wonder if this situation is because:
a) it is not possible for some (not obvious) statistical reasons
b) it is not implemented
c) it is implemented but reporting is suppressed for some reason
d) an unfavoured case of a b.u.g.

I couldn't find any warning in the manual regarding this. Weights are
particularly described on page 433 of the printed manual [ST] (version
10.0). And pweights were specifically mentioned as supported.
Below is an example of why I think d) might be the case (because SEs
are computed in Stata 9, but not in Stata 10).
The reason I think of a) is that if I try to access SEs via
sts generate newvar = se(s)
both Stata 9 and Stata 10 tell me
se(s) not possible with pweighted data
r(404);

I would appreciate if someone could explain how to get SEs of the
survival function in a dataset with pweights and stratification if it
is possible in Stata 10?
It seems to be possible, as mentioned, say, here (Analyzing Health
Equity Using Household Survey Data, A guide to tehniques and their
implementation, page 33 (46 of 234)):
http://www.dcp2.org/file/146/Analyzing%20Health%20Equity%20Using%20Household%20Survey%20Data.pdf
but in light that Stata now does not report them, should the old
estimates be revised?

Thank you,
  Sergiy Radyakin



Example:

use http://www.stata-press.com/data/r9/stan3, clear
sts list, at(1/3)
generate byte w=1
stset t1, id(id) failure(died), [pw=w]
sts list, at(1/3)

******************************************************************************************************
in Stata 9 the last output is

. sts list, at(1/3)

        failure _d:  died
  analysis time _t:  t1
                id:  id
            weight:  [pweight=w]

             Beg.                      Survivor      Std.
   Time     Total     Fail             Function     Error     [95% Conf. Int.]
-------------------------------------------------------------------------------
      1       103        1              0.9903    0.0097     0.9331    0.9986
      2       102        3              0.9612    0.0190     0.8998    0.9852
      3        99        3              0.9320    0.0248     0.8627    0.9670
-------------------------------------------------------------------------------
Note:  Survivor function is calculated over full data and evaluated at
      indicated times; it is not calculated from aggregates shown at left.

******************************************************************************************************
in Stata 10 the last output is:

. sts list, at(1/3)

        failure _d:  died
  analysis time _t:  t1
                id:  id
            weight:  [pweight=w]

             Beg.                      Survivor
   Time     Total     Fail             Function
------------------------------------------------
      1       103        1              0.9903
      2       102        3              0.9612
      3        99        3              0.9320
------------------------------------------------
Note:  survivor function is calculated over full data and evaluated at
      indicated times; it is not calculated from aggregates shown at left.
********************************************************************************************************

Versions of Stata (both for Windows):
-- Stata/SE 9.2 for Windows, Born 20 Jul 2007
-- Stata/MP 10.1 for Windows 64-bit x86-64, Born 02 Feb 2009
*
*   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/
*
*   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