Multiphase Survival Estimation - Data Modeling & Collinearity (Stata)

Basti

New Member
Dear Forum Members and Statistics Ballers!

I am currently trying to model a survival function for entrepreneurs, and would like to specifically test the impact of covariates during certain phases of the endeavor. Simply: Does a Covariate have a stronger or weaker effect during any of the phases?

Phase Indicators are dummys, indicating whether or not a phase is active. Other covariates are continuous or categorical and seem to work without the interaction.

My data looks as follows:
Code:
SAMPID _j COV1 COV2 COV3 PHASE1 PHASE2 PHASE3 PHASE4    _d    _t
000001  1   4    2    1    1      0      0      0        0     0
000001  2   4    2    1    0      1      0      0        0     10
000001  3   4    2    1    0      0      1      0        1     20
000001  4   4    2    1    0      0      0      1        0     30
000002  1   7    1    9    1      0      0      0        0     0
000002  2   7    1    9    0      1      0      0        0     15
000002  4   7    1    9    0      0      0      1        0     40
My regression model is this:
Code:
streg COV1 COV2 COV3 c.COV1#b(0).PHASE1 c.COV2#b(0).PHASE1 c.COV3#b(0)PHASE1, dist(weibull)
However, I seem to have made some mistake in the modeling process, as I get this output:

Code:
------------------------------------------------------------------------------
_t | Haz. Ratio   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
COV1  |   .9743074     .00669    -3.79   0.000      .961283    .9875083
COV2  |   1.031272    1.07964     0.03   0.977     .1325103    8.025958
COV3  |   .4930461   .2154297    -1.62   0.106     .2093952    1.160936
|
COV1#   |
PHASE1  |   1.023131   428.6583     0.00   1.000            0           .
COV2#   |
PHASE1  |   1.023222   199.8965     0.00   1.000     5.2e-167    2.0e+166
COV3#   |
PHASE1  |   .9726358   126.1677    -0.00   1.000     3.7e-111    2.5e+110

This means none of my interaction covariates captures anything. (I know that some of my direct effect covariates are extremely insignificant, but that is ok …)

Can someone please point out to me where exactly my problem is, and how I might be able to model a multiphase parametric survival model? I have until now been unsuccessful in finding such information.

Thank you very much!
Basti

Last edited: