Regression and hazard analysis

[Sheikh Arman Tamim]

Hello All,

I am trying to analyze the impact of an asset transfer program on mortality. This is a randomized control trial study. The intervention took place in 2007 and the endline survey was constructed at 2014. I ran this following OLS regression and found the treatment effect to be negative (that is probability of being dead is less in the treatment group).

areg mortality treatment, absorb(subdistrict) cluster(branchid)

Linear regression, absorbing indicators             Number of obs     = 21,082
Absorbed variable: subdistrict                      No. of categories =     20
                                                    F(1, 39)          =   5.26
                                                    Prob > F          = 0.0272
                                                    R-squared         = 0.0018
                                                    Adj R-squared     = 0.0009
                                                    Root MSE          = 0.1961

                              (Std. err. adjusted for 40 clusters in branchid)
------------------------------------------------------------------------------
             |               Robust
   mortality | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
   treatment |  -.0069882   .0030458    -2.29   0.027    -.0131489   -.0008276
       _cons |   .0444266   .0023893    18.59   0.000     .0395938    .0492594
------------------------------------------------------------------------------

I have also run a similar regression for specific age groups. which also indicates the same.


Then, I tried to do a hazard analysis using the following code,

stset age, fail(mortality==1)
sts list, by(treatment) cumhaz

sts graph, xlabel(0 (5) 70, alternate) by(treatment) cumhaz

I am getting this hazard graph,


Am I doing it correctly?