We'll use KM method in PROC LIFETEST to do survival analysis on a subscribers data set, which looks like this:
censored | start_date | stop_date | [other columns]
---------+------------+------------+----------------
0 | 2001-01-01 | 2002-05-01 |
In this notebook, we'll do these:
PROC LIFETESTKMSurvival is measured.
/* assign datapath pointed to SAS local directory */
libname datapath '/folders/myfolders/';
raw_data¶proc import datafile="/folders/myfolders/censored_start_stop.txt"
out=raw_data dbms=tab replace;
getnames=yes;
run;
* Take a look at the data;
proc sql outobs=5;
select * from raw_data;
quit;
PROC LIFETEST¶The data should be like this:
censored | tenure | [other columns]
---------+--------+----------------
1 | 150 |%macro process_data_for_lifetest(input_data, out_data, cutoff);
/**********************************************************
* Prepare data for PROC LIFETEST
*
* Parameters:
* ----------
* input_data : SAS data set that will be processed
* out_data : SAS data set for PROC LIFETEST
* cutoff : cutoff date for input_data
*********************************************************/
proc sql;
create table &out_data as
select market, channel, censored, start_date, stop_date,
case when censored = 0 and stop_date <= &cutoff
then (stop_date - start_date)
else &cutoff - start_date
end as tenure,
case when censored = 0 and stop_date > &cutoff
then 1
else censored
end as status
from &input_data
where start_date <= &cutoff;
quit;
%mend;
%process_data_for_lifetest(raw_data, tidy_data, '28Dec2008'd)
* Take a look at the tidy data;
proc sql outobs=5;
select * from tidy_data
order by tenure desc;
quit;
KMSurvival is measured.ods listing;
proc lifetest data=tidy_data
method=KM /* use KM estimator */
plots=(survival(nocensor)) /* plot survival curve */
outs=datapath.survival /* output survival estimates */
notable;
time tenure * status(1); /* status: 1-censored */
*strata market;
*strata market channel;
run;
The LIFETEST Procedure
The SAS System 00:02 Monday, November 21, 2016 1
The LIFETEST Procedure
Summary of the Number of Censored and Uncensored Values
Percent
Total Failed Censored Censored
713 255 458 64.24
| Summary of the Number of Censored and Uncensored Values | |||
|---|---|---|---|
| Total | Failed | Censored | Percent Censored |
| 713 | 255 | 458 | 64.24 |
* Take a look at the output data set of survival estimates;
proc sql outobs=5;
select * from datapath.survival;
quit;