Re: st: RE: Multilevel survival analysis

[Steven Samuels <sjsamuels@gmail.com>]

Poisson regression is equivalent to fitting an exponential survival  
distribution by maximum likelihood. The exponential distribution is  
the most unrealistic of distributions, for it assumes that the  
baseline hazard function is constant in time.  To fit more believable  
distributions, see Chapter 7 of the GLLAMM manual.

Steve
sjsamuels@gmail.com

On Feb 15, 2011, at 2:12 PM, Katie Brooks Biello wrote:

Maria, thank you. I have already looked into gllamm and will continue  
to consider it.

Amir, thanks for these suggestions. I just discovered the shared  
option. This would allow me to include a random intercept but not any  
other random effects. This may be sufficient but I'm not quite sure  
yet. I think the Poisson model might be the next best thing if I  
decide that I need to incorporate random effects beyond the random  
intercept. Just to clarify - translating to Poisson from survival  
terminology, the dependent variable would be my censor variable and  
the offset would be my time variable (or log of time), right?

Thanks again.

Katie Brooks Biello, MPH
PhD Candidate
Yale University
School of Epidemiology and Public Health




On Feb 14, 2011, at 2:47 PM, Figueroa Armijos, Maria Augusta (MU- 
Student) wrote:

> Hi Katie,
>
> I am not an expert on survival analysis, but I can help a bit from  
> one experience I had on a project with survival analysis in Stata  
> 11. In the end, I gave up because even though my dataset was big  
> (12,000 obs), it had too many missing values and most of the  
> variables were dummies. I couldn't use HLM because of the missing  
> values, so I tried Stata.
>
> I used xt set and then xtmelogit, options. This command took ages to  
> perform because of my missing values and dummy variables. The  
> longest time I waited was 9 hours. It ran more than 16,000  
> iterations. If you have continuous variables, you could get it to run.
>
> Because xtmelogit didn't work for me, I used -gllam- (user created) http://www.gllamm.org/ 
>  This one worked, and I got the results I needed. The website has  
> books and tutorials you can take.
>
> Good luck!
> Maria
> PhD Student
> Community Policy Analysis Center (CPAC)
> Truman School of Public Affairs
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu 
> ] On Behalf Of Katie Brooks Biello
> Sent: Monday, February 14, 2011 1:30 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: Multilevel survival analysis
>
> Hi -
> I am using Stata/SE 11.1 for Windows (32-bit). I have multilevel data
> where individuals are nested in metropolitan areas (MAs). I have a
> time-to-event outcome, and want to estimate the effect of an MA-level
> variable on this time-to-event outcome, controlling for other MA-level
> variables and individual-level variables. In other words, I am hoping
> to do the survival equivalent of HLM. It is my understanding that
> Stata 11 can perform multilevel survival analysis but I can not find
> direct guidance on what commands are used and how this is performed.
> Does anyone have experience with this? Can you offer up the code
> necessary to run this?
>
> Thank you,
>
> Katie Brooks Biello, MPH
> PhD Candidate
> Yale University
> School of Epidemiology and Public Health
>
>
>
>
> *
> *   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/


*
*   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/