Help the Stat Consulting Group by giving a gift

**Stata Textbook Examples
Applied Survival Analysis**

**Please Note:** Stata **graph** commands changed with version 8 and
this page was developed before version 8 was released and uses Stata 7 **graph**
commands. Please see
How do I use version 7 graph commands in Stata version 8? for
information on how to either run these Stata 7 **graph** commands
in Stata version 8, or how you can covert these commands to use Stata 8
syntax.

Let's first discussed the usage of categorical time-dependent variables.
These variables are extremely useful for multi-state models. An example of
a multi-stage could be when subjects are in or out of a hospital or other
treatment programs. Thus, we would like to have a variable which would be
equal to one when the subject is in the hospital and is equal to zero when the
subject has been discharged from the hospital.

To illustrate how to generate a categorical, in this instance a dichotomous,
time-dependent variable we will use the model shown in table 7.3 on p. 252 in
the book "Applied Survival Analysis" by Hosmer and Lemeshow. This model
includes a time-dependent variable called **off_trt**. The data set
that we'll be using is the **UIS** data set and a description of the data can
be found on p. 23 of the book. The main focus of the model is to predict
the length of time until a subject starts using drugs again. The main
variables that we will consider are **time**, **los** and **censor**.
The variable **time** is the total length of time from admission until the
subject was no longer in the study either because the subject returned to drug
use or the subject was lost to follow up (censored). The variable **los**
measures the length of time from admission until the subject left the treatment
program. The variable **censor** indicates whether the subject returned to
drug use (**censor**=1) or was lost to follow up (**censor**=0).

Suppose we wanted a variable called
**off_trt** that would be an indicator
that the subject has left the treatment program. The variable **off_trt**
would equal zero if **time** <= **los** and
equal 1 if **time** > **los**.

use uis, clear gen id = _n

It is actually really straightforward in Stata to create such a variable. The
command **stsplit** is all we need here.

Then we create the variables need in the model in table 5.11, p. 252.stset time, f(censor) id(ID) stsplit off_trt, at(0) after(_t=los) replace off_trt = off_trt + 1 list id los time _t _t0 _d off_trt in 45/55, clean nolid los time _t _t0 _d off_trt 45. 24 48 48 48 0 0 0 46. 24 48 53 53 48 1 1 47. 25 90 90 90 0 0 0 48. 25 90 225 225 90 1 1 49. 26 91 91 91 0 0 0 50. 26 91 161 161 91 1 1 51. 27 87 87 87 0 1 0 52. 28 88 88 88 0 0 0 53. 28 88 89 89 88 1 1 54. 29 9 9 9 0 0 0 55. 29 9 44 44 9 1 1

gen ivhx3 = ivhx==3 gen agsi = age*site gen rasi = race*site fracgen ndrugtx -1 -1 rename ndrugt_1 fp1 rename ndrugt_2 fp2

stcox age becktota fp1 fp2 ivhx3 race treat site agsi rasi off_trt, nohr

No. of subjects = 575 Number of obs = 1079 No. of failures = 464 Time at risk = 138900 LR chi2(11) = 404.66 Log likelihood = -2461.6565 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ _t | _d | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- age | -.037877 .0100552 -3.77 0.000 -.0575849 -.0181691 becktota | .0079745 .0049148 1.62 0.105 -.0016584 .0176073 fp1 | -.6086314 .1283462 -4.74 0.000 -.8601853 -.3570775 fp2 | -.2255842 .0495962 -4.55 0.000 -.322791 -.1283774 ivhx3 | .2746718 .1089458 2.52 0.012 .0611419 .4882016 race | -.5169569 .1344993 -3.84 0.000 -.7805707 -.2533432 treat | .0193972 .0961311 0.20 0.840 -.1690164 .2078108 site | -.9690888 .5158775 -1.88 0.060 -1.98019 .0420125 agsi | .0363502 .0158037 2.30 0.021 .0053755 .0673249 rasi | .5109338 .256902 1.99 0.047 .0074152 1.014452 off_trt | 2.571152 .1567619 16.40 0.000 2.263904 2.8784 ------------------------------------------------------------------------------

The content of this web site should not be construed as an endorsement of any particular web site, book, or software product by the University of California.