Wednesday, March 20, 2013

SAS: Adding Watermarks to ODS Report - LIFETEST, LIFEREG, and PHREG

Building off the Weibull survival simulation, and a quick snippet of code to create a watermark background, we can create a quick report with SAS ODS. If you want to play along at home, just add the following code to the survival simulation: it will seed the random number generator so you should get the exact same data:

set.seed(1)

Now, let's start up a report with SAS 9.2 using ODS: we can load in the data using IMPORT, and start playing around with LIFETEST, LIFEREG, and PHREG. This won't be extensive or exhaustive- just an excuse to show off a new trick. If you're new to survival analysis in SAS, or survival analysis in general, there's plenty of material over at UCLA's Institute for Digital Research and Education- possibly one of the best stat resources out there.

/* This will keep the watermark properly aligned*/
options papersize = (8.5in 11in);

proc import
    datafile = "C:\Data\Didactic\Data\Survival sim 2013-03-20.csv"
    out = surv dbms = csv replace;
run;


/* create our template using the results */
proc template;
    define style watermark;
    parent = styles.printer;
    style header from header / background=_undef_;
    style body from document / background=_undef_
        backgroundimage = "C:\Data\Didactic\Results\2013-03-20\wm - CONFIDENTIAL 03-19-2013 300 dpi.png";
    end;
run;


/* begin the PDF report */
ods pdf file = "C:\Data\Didactic\Results\2013-03-20\Survival Sim - 2013-03-05.pdf"
    style = watermark;
ods noproctitle;
ods escapechar="^";
ods pdf text = "^S={just = center font_weight = bold font_style = italic
    font_size = 18pt}Survival Simulated Data";
 

/* plot survival, log survival, and log-log survival */
ods proclabel = "Survival - X1";
title "Survival Function - Stratified by X1";
proc lifetest data = surv plots = (s, ls, lls) graphics notable;
    time time_to_event*event(0);
    strata x1;
run;

ods proclabel = "Survival - X2";
title "Survival Function - Stratified by X2";
proc lifetest data = surv plots = (s, ls, lls) graphics notable;
    time time_to_event*event(0);
    strata x2;
run;

ods proclabel = "Survival - X3";
title "Survival Function - Stratified by X3";
proc lifetest data = surv plots = (s, ls, lls) graphics notable;
    time time_to_event*event(0);
    strata x3;
run;


ods proclabel = "Weibull GLM";
title "Parametric Regression - Weibull";
proc lifereg data = surv;
    class x1 x2 x3;
    model time_to_event*event(0) = x1 x2 x3 x4 x5 x6
        x7 x8 x9 x10 x11 / dist = weibull;
run;

ods proclabel = "Cox PH Model";
title "Nonparametric Regression - Cox Proportional Hazards";
proc phreg data = surv;
    class x1 x2 x3;
    model time_to_event*event(0) = x1 x2 x3
        x4 x5 x6 x7 x8 x9 x10 x11;
run;
ods pdf close;


The results aren't too shabby, and should give you something like this:


Another day, we can delve deeper into the dataset in SAS, or maybe boot up the Survival package in R.

No comments:

Post a Comment