Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: hshaz interpretation of the unobserved heterogeneity parameters

 From From Judit VALL CASTELLO To Subject st: hshaz interpretation of the unobserved heterogeneity parameters Date Thu, 29 Sep 2011 10:40:03 +0100 Date Wed, 28 Sep 2011 12:05:38 +0200 (CEST)

```parameters

Dear all,

I am fitting hshaz for the first time and I have problems with
interpreting the coefficients and the unobserved heterogeneity
parameters.
This is the output that I get fitting a model with flexible baseline
hazard for each year (15 years) and ommiting one of the dummies instead
of the constant term (only showing the estimation of the model with UH):

<snip>

Discrete time PH model, with discrete mixture     Number of obs   =
101139
LR chi2()       =
.
Log likelihood = -13131.227                       Prob > chi2     =
.

-
------------------------------------------------------------------------
------
employment |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
Interval]
-
-------------+----------------------------------------------------------
------
hazard       |
d2 |   2.202138   .0835986    26.34   0.000     2.038287
2.365988
d3 |   1.817322   .0942435    19.28   0.000     1.632608
2.002036
d4 |   1.559679   .1042669    14.96   0.000      1.35532
1.764039
d5 |   1.326683   .1144048    11.60   0.000     1.102453
1.550912
d6 |   .9699759   .1286689     7.54   0.000     .7177896
1.222162
d7 |   .9914302   .1361961     7.28   0.000     .7244907
1.25837
d8 |   .5558144   .1620074     3.43   0.001     .2382858
.873343
d9 |   .5627367   .1741448     3.23   0.001     .2214191
.9040543
d10 |   .3581391   .2022719     1.77   0.077    -.0383066
.7545848
d11 |   .3722488   .2234632     1.67   0.096    -.0657311
.8102287
d12 |   .4746029   .2471491     1.92   0.055    -.0098004
.9590061
d13 |   .3426948   .3125202     1.10   0.273    -.2698336
.9552232
d14 |  -.5513795   .5904984    -0.93   0.350    -1.708735
.6059762
d15 |   .3091984   .5912455     0.52   0.601    -.8496214
1.468018
fem |  -.6304255   .0524701   -12.01   0.000     -.733265
-.5275859
agedisabil~y |  -.0880126   .0030303   -29.04   0.000    -.0939519
-.0820732
totaldis |  -3.095721   .2213199   -13.99   0.000      -3.5295
-2.661942
profcateg2 |   .1461056   .0541343     2.70   0.007     .0400044
.2522068
profcateg3 |   .1733369   .0991126     1.75   0.080    -.0209203
.3675941
lnbasereg |   .2052545   .0567508     3.62   0.000      .094025
.3164839
pensln |  -.3220117   .0415467    -7.75   0.000    -.4034417
-.2405816
selflj2 |  -.0656422   .0537471    -1.22   0.222    -.1709846
.0397002
totempspe |   .0086617   .0009305     9.31   0.000      .006838
.0104854
ur |   -.033141   .0036001    -9.21   0.000    -.0401971
-.026085
_cons |   1.568922   .5036847     3.11   0.002     .5817179
2.556126
-
-------------+----------------------------------------------------------
------
m2           |
_cons |    .920075   .2086793     4.41   0.000      .511071
1.329079
-
-------------+----------------------------------------------------------
------
logitp2      |
_cons |  -.1763005   .8397107    -0.21   0.834    -1.822103
1.469502
-
-------------+----------------------------------------------------------
------
Prob. Type 1 |   .5439613   .2083048     2.61   0.009     .1870183
.8608183
Prob. Type 2 |   .4560387   .2083048     2.19   0.029     .1391817
.8129817
-
------------------------------------------------------------------------
------
Note: m1 = 0

The prob.Type 1 and 2 is the probability that someone in my sample
belogs to the first (or second) type of individuals, but how should I
interpret the m2 and the logitp2?

Also, how can I translate the coefficients (for example of the ur) into
marginal effects or probabilities?

====

-hshaz- is on SSC. Please mention the provenance of user-written
commands.

The information you seek is provided in the help file. logitp2 =
logit(prob type 2). m2 is the value of mass point 2 (mass point 1 is
normalised at zero.)

This is a proportional hazards model (applied to
interval-censored/discrete time data), so interpret coefficients in the
standard PH fashion.

Predictions of all kinds, including survival probabilities, require some
assumption about which 'class' a person belongs to (unless you
-predict-.

You could also read the Survival Analysis Using Stata web material (URL
below), including the Lesson which illustrates these models in action.

Stephen
------------------
Professor Stephen P. Jenkins <s.jenkins@lse.ac.uk>
Department of Social Policy and STICERD
London School of Economics and Political Science
Houghton Street, London WC2A 2AE, UK
Tel: +44(0)20 7955 6527
Changing Fortunes: Income Mobility and Poverty Dynamics in Britain, OUP
2011, http://ukcatalogue.oup.com/product/9780199226436.do
Survival Analysis Using Stata:
http://www.iser.essex.ac.uk/survival-analysis

Please access the attached hyperlink for an important electronic communications disclaimer: http://lse.ac.uk/emailDisclaimer

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```