Re: st: Query re. stsplit and survival modelling

[Steve Samuels <sjsamuels@gmail.com>]

Eleonara,

The Statalist FAQ ask that you provide code that shows what you did. I don't understand what you mean by splitting "episodes".  

But to answer your question in general:

1) With time-dependent variables in -stcox-, you need split only "at(failures)"

2)  For a Weibull parametric model with time-dependent variables, you
need to split at the months that values change. 

There is no reason for either model to split any further.

Steve

On Dec 13, 2012, at 10:34 AM, Eleonora Paesen wrote:

Hello,
I am using a dataset that includes monthly observations concerning
several variables.
I have used stsplit in order to include time-varying variables into
Cox proportional and Weibull hazard models. Before running the models
I have made sure that all time-varying variables has been opportunely
stsplitted.
However I have not expanded the dataset in order to have person-month
observations by using the following command line:
stsplit time, at(1(1)n)
where 'n' corresponds to the 'last observed exit t' obtained after
stsetting the data.
Does anyone could explain the value of the expansion of the data on a
strictly monthly basis instead of 'splitted episodes' basis, whose
length could be more than one month?
How the coefficients obtained in the models runned on the two datasets
differ? Is there a practical reason for splitting the data on a
monthly basis?
Running times increase enormously and I am not sure there is any gain
by st-splitting the data in order to obtain person-months
observations.
Any help or suggestionson the matter would be greatly appreciated.
Best,
Eleonora
*
*   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/