Hi everyone, i'm working on a dataset with counting process structure. I have something like this:
Id Status Event tstart tstop sus
1 1 1 0 16377 cocaine
1 0 2 18019 18213 alcohol
2 0 1 0 18169 alcohol
3 1 1 0 16037 cocaine paste
3 1 2 16049 16129 cocaine paste
3 0 3 16212 16275 cocaine paste
4 0 1 0 16617 cocaine
5 1 1 0 16750 marijuana
5 0 2 17123 17525 marijuana
Notice that the event is hospital readmission, so i thought a PWP-TT was a good choice. I tested de PH assumption using this code:
Code:
stset tstop, fail(status) exit(time .) id(id) enter(tstart)
stpm2 sus, df(5) scale(hazard)
predict su1, at(sus 1) hazard
predict su2, at(sus 2) hazard
predict su3, at(sus 3) hazard
predict su4, at(sus 4) hazard
predict su5, at(sus 5) hazard
stphplot, by(sus)
stcox i.sus
stphtest, rank detail
Code:
stphtest, rank detail

      Test of proportional-hazards assumption

      Time:  Rank(t)
      ----------------------------------------------------------------
                  |       rho            chi2       df       Prob>chi2
      ------------+---------------------------------------------------
      1b.sus      |            .            .        1             .
      2.sus       |     -0.05253        66.45        1         0.0000
      3.sus       |     -0.08080       156.46        1         0.0000
      4.sus       |     -0.04212        42.66        1         0.0000
      5.sus       |     -0.10743       276.38        1         0.0000
      ------------+---------------------------------------------------
      global test |                    341.50        4         0.0000
      ----------------------------------------------------------------
The results suggest that substance has a time-dependent effect, so i tried this:
Code:
stcox sus, tvc(sus) strata(event)
Code:
Stratified Cox regr. -- Breslow method for ties

No. of subjects =       79,710                  Number of obs    =     103,775
No. of failures =       24,066
Time at risk    =   1338295692
                                                LR chi2(2)       =     2273.74
Log likelihood  =   -239219.92                  Prob > chi2      =      0.0000

------------------------------------------------------------------------------
          _t | Haz. Ratio   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
main         |
         sus |   3.071586    .211246    16.32   0.000     2.684244    3.514823
-------------+----------------------------------------------------------------
tvc          |
         sus |   .9999414   4.21e-06   -13.90   0.000     .9999332    .9999497
------------------------------------------------------------------------------
                                                           Stratified by event
Note: Variables in tvc equation interacted with _t.
The question is if its possible to plot the baseline hazard for the strata, taking into account the TD effects?

i found solutions for td effects like
Code:
scurve_tvc
but i dont know how to plot strata and td effects jointly, it is possible? it makes sense?


Thank you for your help