*******************
* duration model duratio.do *
*******************

* In stata, you don't have to take log explicitly. Constant terms is implicitly assumed.
* log-linear model for strike duration

log using duration.log, replace

insheet using duration.txt, clear
* declare t as survival time data
stset t


* graph the Kaplan-Meier survivor function
sts graph

* graph the Nelson-Aalen cumulative hazard function
sts graph, na

* Cox's proportional Hazard model for T with covariates 
stset t
stcox prod, nohr

*  log-linear parametric hazard model
stset t
streg prod, distribution(weibull) time

* the regression above is accelerated failure-time form. Following regression is log relative-hazard form
streg prod, distribution(weibull) 

* to specify that coefficients rather than exponentiated coefficients are to be displayed 
streg prod, distribution(weibull) nohr

* exponential distribution
streg prod, distribution(exponential) time

* log-normal distribution, the option time is disabled, so this is a log relative-hazard form
streg prod, distribution(logn) 

* log-logistic distribution, the option time is disabled, so this is a log relative-hazard form
streg prod, distribution(logl) 

* Same results are yielded as those of limdep
* Note: Limdep makes no explicit distinction between accelerated failure-time form and log relative-hazard form
* the results produced by Dr Lee's limdep codes amounts stata command without option "nohr"