parametric survival regression and discrete time survival regression

[shahla ebrahimi]

Dear Statalisters
As I read in this article,

1- for time-dependent covariates there are no need to test any violation of the proportional hazards assumptions (page-7).

2- natural logarithms of firms' annual age are used as baseline hazard function (page-23). Could you please let me know what is the difference between panel logistic regression and discrete time survival analysis except including a baseline hazard?
Really, If I include time dummy variables (12 dummy variables) in the model, it would take a long time to implement or it reports not concave.

3- it uses random effects because some firms are more ‘frail’ and thus have a higher likelihood of experiencing default (page-8). Could you please let me know if there is no deed to do Hausman, Breusch-Pagan, Chow tests in order to determine fixed effects vs random effects should be used?

4- Could you please let me know why in Stata the result of implementing random effects parametric survival model is different from that of implementing random effects panel logistic? I know that the former reports hazard rate but the latter reports odds ratio. However, the significance of predictors also are different.
a- statistics-longitudinal/panel data-binary outcomes-logistic regression
b- statistics-longitudinal/panel data-parametric survival regression.
Is it necessary for both to include baseline hazard function (ln(time))?
Thanks in advance.
Best regards,

[Maarten Buis]

With such things the devil is in the detail. To get a more useful answer tell us about the actual model that you try to estimate, the actual data you want to use, how you try to implement that, and the questions that come up.

If you find your current source of information about survival analysis confusing then you should try another source. I like (Cleves, Gould, Marchenko 2016) and (Allison 2014).

Allison, P.D., 2014. Event History and Survival Analysis: Regression for Longitudinal Event Data. 2nd ed. Thousand Oaks: Sage.

Mario Cleves, William W. Gould, and Yulia V. Marchenko (2016) An Introduction to Survival Analysis Using Stata, Revised Third Edition. College Station: Stata Press.

[shahla ebrahimi]

Hi. Thanks for your useful information and suggestion. In fact, I want to use discrete time survival analysis for bankruptcy problem. However, the more I read the more I am confused so I don't know how to estimate. What tests I should do. Unfortunately, I could not find an example with all required tests and details.
Please kindly find some part of my data attached.
I want to estimate this model:
Event=x1+x2+x3+x4+x5
d1 to d12: dummy variables for each time period.
Thanks in advance.
Best regards,