Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: RE: - svy: stcox - (Stata 11)


From   Steve Samuels <[email protected]>
To   [email protected]
Subject   Re: st: RE: - svy: stcox - (Stata 11)
Date   Sun, 21 Jul 2013 17:36:30 -0400

I'd only add: 

• With such strong predictors, I'd be surprised if hazards were not
proportional for some categories.

• If you are studying survival in a population, I'd like to know why the
time scale is other than age. Perhaps time is defined by entry into some
kind of program. In any case, please describe the study in more detail.

S.


Use -stcurve- to estimate the survival rates; output them to new data sets.

1 & 2: Compute cumulative mortality rates from the survival rates. Then plot and summarize.

3: Expected life is the integral of the survival curve (see any survival
text, including p. 48 of
https://www.iser.essex.ac.uk/files/teaching/stephenj/ec968/pdfs/
ec968lnotesv6.pdf. Therefore, you could apply -integ- to each curve in
2.  If the survival curves don't descend to zero or to a "small"
number>0, then the the computed expectation will be less than the true
expectation.

If this situation arises, you can compute a restricted mean by first
adding a fake data point with time ≥ the highest censored values and
survival equal to zero. An easier approach might be to create a hazard
model with Paul Lambert's -stpm2- command (SSC), followed by -predict-
with the restricted mean options.

Some comments on your model:

1. I don't like using "missing" as a predictive category with survey data. I think that multiple imputation is the preferred approach in this situation.
2. Are there no interactions, no calendar time influences?


Steve
[email protected]






On Jul 19, 2013, at 4:30 PM, Muhuri, Pradip (SAMHSA/CBHSQ) wrote:

Hi,

Using  - svy: stcox- and then post estimation commands (for which I need your help on), my goals are to calculate the following:

	(1) the baseline mortality rate (for persons who are classified in the reference category 	for each covariate in the model);

	(2) the mortality rate for particular  combination of the levels of covariates 	(e.g.,  xspd2=1, agegrp5=5 & sex=2); and

	(3) model-based life expectancy for particular  combination of the levels of covariates 	(e.g., xspd2=1, agegrp5=5  & sex=2).

The following code gives me the output that is limited to coefficient, etc.

     svyset psu [pweight=wt8], strata (stratum) vce(linearized) singleunit(missing)
     stset interval [pweight=wt8], failure(mortstat==1)  
     capture noisily: svy: stcox  ib2.xspd2  i.agegrp5 i.sex  ib3.x_xsmoke, nohr

Survey: Cox regression

Number of strata   =       339                 Number of obs      =     238811
Number of PSUs     =       678                 Population size    =  200969312
                                             Design df          =        339
                                             F(   9,    331)    =    2090.43
                                             Prob > F           =     0.0000

------------------------------------------------------------------------------
           |             Linearized
        _t | Haz. Ratio   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   1.xspd2 |   2.071856   .0856399    17.62   0.000     1.910069    2.247346
           |
   agegrp5 |
        2  |   3.290646   .1558252    25.15   0.000     2.997982    3.611881
        3  |    8.49614   .3434431    52.93   0.000     7.846751    9.199271
        4  |   21.41636   .7986149    82.17   0.000     19.90173    23.04627
        5  |   64.71143   2.359975   114.34   0.000     60.23198    69.52402
           |
     2.sex |   .7195701   .0139644   -16.96   0.000     .6926199    .7475689
           |
  x_xsmoke |
        1  |   2.233029   .0573213    31.30   0.000     2.123078    2.348674
        2  |    1.32416   .0291701    12.75   0.000     1.268008    1.382798
        9  |   1.464324   .1946862     2.87   0.004     1.127359    1.902009
------------------------------------------------------------------------------

Could anyone please help me with the Stata postestimation code and point me to the relevant references to produce results for (1), (2), and (3), as referred to above?

Thank you in advance.

Pradip

Pradip K. Muhuri, PhD
Statistician
Substance Abuse & Mental Health Services Administration
The Center for Behavioral Health Statistics and Quality
Division of Population Surveys
1 Choke Cherry Road, Room 2-1071
Rockville, MD 20857

Tel: 240-276-1070
Fax: 240-276-1260
e-mail: [email protected]

The Center for Behavioral Health Statistics and Quality your feedback.  Please click on the following link to complete a brief customer survey:   http://cbhsqsurvey.samhsa.gov


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


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index