Hello SAS experts,
I'm trying to run following codes for relative survival and hit following error. Any idea why? I appreciate your taking your precious time out for help.
date sources (colon and popmort) and lexis.sas can be found here: http://www.pauldickman.com/survival/?dir=sas
872 proc summary data=&individ nway;
873 var d w p_star y d_star sex;
874 id range length;
875 class &vars fu; /* Follow-up must be the last variable in this list */
876 output out=&grouped(drop=_type_ rename=(_freq_=l)) sum(d w y d_star)=d w y d_star mean(p_star)=p_star;
877 format &formats ;
ERROR: The format SEX was not found or could not be loaded.
878 run;
NOTE: The SAS System stopped processing this step because of errors.
WARNING: The data set COLON.GROUPED may be incomplete. When this step was stopped there were 0 observations and 0 variables.
WARNING: Data set COLON.GROUPED was not replaced because this step was stopped.
NOTE: PROCEDURE SUMMARY used (Total process time):
real time 0.39 seconds
cpu time 0.01 seconds
LIBNAME colon "D:\";
/****************************************************************
SURVIVAL.SAS
Estimate relative survival and produce output data files that
can be used for fitting relative survival models.
http://www.pauldickman.com/rsmodel/
Paul Dickman (paul.dickman@ki.se)
Version 1.0 - June 2004
****************************************************************/
title; footnote;
title1 'Colon carcinoma diagnosed in Finland 1975-1994 (follow-up to 1995)';
options fmtsearch=(colon work library) orientation=landscape pageno=1;
/****************************************************************
Define the input and output files.
****************************************************************/
/* Population mortality file */
%let popmort=colon.popmort ;
/* Patient data file */
%let patdata=colon.colon ;
/* Output data file containing individual records */
%let individ=colon.individ ;
/* Output data file containing collapsed data */
%let grouped=colon.grouped ;
/****************************************************************
The macro variable VARS carries the variables over which the
life tables are stratified.
For example:
%let vars = sex yydx age;
Will result in a lifetable being estimated for each combination
of sex, yydx (year of diagnosis), and age. If, for example, the
variable age contains age at diagnosis in years then categories
can be constructed using a format.
****************************************************************/
%let vars = sex yydx age;
%let formats = sex sex. age age. yydx yydx. ;
data &individ;
length id 5;
set &patdata;
/* Restrict to localised stage*/
if stage=1;
/* Create a unique ID for each individual */
id+1;
/****************************************************************
The variable SURV_MM contains survival time in completed months.
We will add 0.5 to all survival times, both to avoid problems
with individuals with time=0 (who are theoretically never at risk
and may be excluded from some analyses) and because this provides
a more accurate estimate of person-time at risk (for Poisson
regression analyses).
****************************************************************/
surv_mm = surv_mm + 0.5;
/* The lexis macro requires a variable containing the time at entry */
entry=0;
/* Create an indicator variable for death due to any cause */
if status in (1,2) then d=1;
else d=0;
drop stage subsite status;
label id='Unique subject ID';
run;
/*****************************************************************
It is preferable to put the lexis macro in the autocall library
rather than including it as is done here
****************************************************************/
%include 'D:....\lexis.sas';
/*****************************************************************
Split the data to obtain one obervation for each life table interval
for each individual. The scale must be transformed to years.
****************************************************************/
%lexis (
data=&individ.,
out=&individ.,
breaks = %str( 0 to 10 by 1 ),
origin = 0,
entry = entry,
exit = surv_mm,
fail = d,
scale = 12,
right = right,
risk = y,
lrisk = ln_y,
lint = length,
cint = w,
nint = fu)
;
/****************************************************************
Create variables for attained age and calendar
year which are 'updated' for each observation for a single
individual. These are the variables by which we will merge in
the expected probabilities of death, so they must have the
same names and same format as the variables indexing the
POPMORT file (sex, _age, _year in this example).
****************************************************************/
data &individ;
set &individ;
/*********************************************************************
Create a variable for attained age at the start of the interval).
This variable must have the same name and have the same format as the
corresponing variable in the popmort file.
*********************************************************************/
_age=floor(age+left);
/*******************************************************************
Create a variable for calendar period at the start of the interval.
This variable must have the same name and have the same format as the
corresponing variable in the popmort file.
*******************************************************************/
_year=floor(yydx+left);
/* A variable to label the life table output */
range=put(left,4.1) || ' - ' || left(put(right,4.1));
drop entry left right;
run;
/****************************************************************
Now merge in the expected probabilities of death.
****************************************************************/
proc sort data=&individ;
by sex _year _age;
run;
proc sort data=&popmort;
by sex _year _age;
run;
data &individ;
length d w fu 4 y ln_y length 5;
merge &individ(in=a) &popmort(in=b);
by sex _year _age;
if a;
/* Need to adjust for interval lengths other than 1 year */
p_star=prob**length;
/* Expected number of deaths */
d_star=-log(p_star)*(y/length);
*keep &vars fu range length d w p_star y ln_y d_star;
label
d_star='Expected number of deaths'
d='Indicator for death during interval'
w='Indicator for censored during interval'
y='Person-time (years) at risk during the interval'
length='Interval length (potential not actual)'
ln_y='ln(person-time at risk)'
p_star='Expected survival probability'
_age='Attained age'
_year='Attained calendar year'
range='Life table interval'
fu='Follow-up interval'
sex='Sex'
;
run;
/****************************************************************
Collapse the data to produce the life table.
****************************************************************/
proc summary data=&individ nway;
var d w p_star y d_star;
id range length;
class &vars fu; /* Follow-up must be the last variable in this list */
output out=&grouped(drop=_type_ rename=(_freq_=l)) sum(d w y d_star)=d w y d_star mean(p_star)=p_star;
format &formats ;
run;
