topic Re: For loops in SAS Programming

Please what is wrong with this macro,why cant I get the data data impute_t_&p below

desireatem — Fri, 08 Nov 2013 02:47:48 GMT

ods html close;

%macro surv(replicates=);

%do p=1 %to &replicates;

data tes_&p;

seed=-1;

alpha1 = 2.50;

beta1 = 2.00;

beta2=4;

do i = 1 to 500;

X=ranbin(0,1,.5);

lambdaT = 0.025; *baseline hazard;

lambdaC= .03; *heavy=0.15; *light=0.03;

er=0+(0.05)*rannor(1);

t = rand("WEIBULL", 0.75, lambdaT); * time of event;

c1 = rand("WEIBULL", 1.25, lambdaC) ;* time of censoring;

c2= rand("WEIBULL", 1.50, lambdaC) ;

if er le 0.240482 then c=c1; else c=c1;

time = min(t, c); * which came first?;

censored = (c lt t);

obs=(t lt c); * creating observation vaiable from censored when observeed obs=1;

obs = (t lt c);

y= alpha1 + beta1*t + er;

if t le 0.015072 then R=0;

else if t gt 0.015072 then R=1;

y1= alpha1 +beta1*R + er;

output;

end;

run;

%do i=1 %to 7;

proc surveyselect data=tes_&i method = urs sampsize = 10 OUTHITS

reps=1 seed=123410 out=test_&i;

run;

proc means data = test_&i sum;

where obs = 0;

var censored;

ods output summary = summary_&i;

run;

%end;

data impute_t_&p;

set summary_1 -summary_&i;;

run;

%end;

%mend surv;

%surv(replicates=10);

Re: For loops

Patrick — Fri, 08 Nov 2013 03:07:36 GMT

You can do what you want but the syntax is different (it looks like SAS and not JMP). You might want to have a read of the SAS Macro Language syntax, e.g. the syntax could be:

/* define and compile the macro */

%macro xxxx; ...%do i=1 %to 7; ... out=test_&i... %end; ...%mend;

/* call the macro */

%xxxx;

Re: For loops

desireatem — Fri, 08 Nov 2013 04:05:11 GMT

So I can have a macro within a macro?

Re: For loops

desireatem — Fri, 08 Nov 2013 04:46:21 GMT

THis is how my code looks like;, is this format correct, why is it that I cant get the data impute_t_&p

ods html close;

%macro surv(replicates=);

%do p=1 %to &replicates;

data tes_&p;

seed=-1;

alpha1 = 2.50;

beta1 = 2.00;

beta2=4;

do i = 1 to 500;

X=ranbin(0,1,.5);

lambdaT = 0.025; *baseline hazard;

lambdaC= .03; *heavy=0.15; *light=0.03;

er=0+(0.05)*rannor(1);

t = rand("WEIBULL", 0.75, lambdaT); * time of event;

c1 = rand("WEIBULL", 1.25, lambdaC) ;* time of censoring;

c2= rand("WEIBULL", 1.50, lambdaC) ;

if er le 0.240482 then c=c1; else c=c1;

time = min(t, c); * which came first?;

censored = (c lt t);

obs=(t lt c); * creating observation vaiable from censored when observeed obs=1;

obs = (t lt c);

y= alpha1 + beta1*t + er;

if t le 0.015072 then R=0;

else if t gt 0.015072 then R=1;

y1= alpha1 +beta1*R + er;

output;

end;

run;

%do i=1 %to 7;

proc surveyselect data=tes_&i method = urs sampsize = 10 OUTHITS

reps=1 seed=123410 out=test_&i;

run;

proc means data = test_&i sum;

where obs = 0;

var censored;

ods output summary = summary_&i;

run;

%end;

data impute_t_&p;

set summary_1 -summary_&i;;

run;

%end;

%mend surv;

%surv(replicates=10);

Re: For loops

Patrick — Fri, 08 Nov 2013 05:37:48 GMT

You can define a macro within a macro but if you want to use the macro then you need to call it. So you would need to add %selecting; to your code.

You normally avoid to define a macro in a macro because in doing so the inner macro definition gets re-compiled for every call of the out macro. So what you would do if such a nesting is required is a construct like:

%macro inner;

....

%mend;

%macro outer;

%inner;

%mend;

%outer;

Also don't forget that macro variables defined within a macro have per default a scope of local. So &i defined in macro %selecting is not available in the outer macro %surv. You need to use a %global i; in your inner macro to change this.

I also believe that at the end of a %do loop the value of &i is the upper boundary of the loop plus 1 - so "summary_&i;" would expect a data set "summary_8" - which doesn't exist.

I actually don't understand why you need macro %selecting at all. Can't you just have your %do loop as part of macro %surv?

Another option to collect data created iteratively would be:

proc append base=impute_t_&p data=summary_&i;

run;

You would have this proc append step inside your do loop. If you don't need to keep a table per iteration then you could also overwrite the summary_.. table in each operation and so using syntax like:

proc append base=impute_t_&p data=summary;

run;

Haven't really looked at the rest of the code. What you normally want to do is make your program work without a macro (so only for a single case) and only once this is working "macrotize" it.

Also: It's most of the times better to use by-group processing instead of calling a procedure multiple times with different data slices. So eventually you could combine the output from proc sureveyselect using proc append but then call proc means only once.