I am trying to do a survival analysis on a data set with an underlying
hierarchy. All data is either left-, right- or interval censored and there
are 3 (or possibly 4) levels of the hierarchy. I would like to fit a
parametric survival model such as Weibull.
What I would like to know is: if I simply wish to determine a best fitting
survival curve without piecewise confidence intervals, do I need to account
for the multilevel structure? My understanding is that the random effects
in a multilevel model are normally distributed with zero mean so the
parameter best estimates will not change whether or not the random effects
are included. Hence I would expect the survival curves from single-level
and multilevel analyses to be the same.
Although I know that failing to account for the multilevel structure can
lead to inaccurate confidence intervals, I am confident that the parameters
I want to include are significantly associated with my response variable,
through previous cross-sectional multilevel generalised linear modelling of
the same data set at various time points. If I don't need to include
random effects I would prefer not to, as the analysis is a lot simpler and
would not require Stata.
However, if I do need to account for the random effects, can such an
analysis be performed in Stata?
Thanks
John
*
* 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/
