Date:         Wed, 18 Oct 2006 09:42:22 -0200
Reply-To:     Renata Matos <renata_matos@YAHOO.COM.BR>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Renata Matos <renata_matos@YAHOO.COM.BR>
Subject:      Re: Macro + Surveyselect
Comments: cc: jim2stat@YAHOO.CO.UK, davidlcassell@MSN.COM
Content-Type: text/plain; charset="iso-8859-1"

Thank you for de messages, Jim Groeneveld and David L. Cassell.

I'll explain what I'm trying to do. There are many steps and in the first one I have to take J samples of a data that may vary. J ranges from ll=0.1(n/k) to lu=0.5(n/k). n varies according to the data and k is defined by the user. I don't know if it's Ok but I put J=mean(ll,lu)+1. That's not the problem. I can fix this latter. I agree with David and I think that IML is not necessary here. I did this using data _null_ and call symput and worked well. But now I've some other questions:

-          I did a macro to discover the number of objects in the data but I think it's stupid. I don't know. Any suggestions?

-          The dataset nobs has the number of objects. And part of the code is:

data _null_;

      set nobs;

      call symput('n',nobs);

run;

data _null_;

      ji=0.1*(&n/&k);

      js=0.5*(&n/&k);

      aux=int((ji+js)/2)+1;

      call symput('quantidade',aux);

run;

Why I can't put this two data _null_ together? Like this:

data _null_;

      set nobs;

      call symput('n',nobs);

      ji=0.1*(&n/&k);

      js=0.5*(&n/&k);

      aux=int((ji+js)/2)+1;

      call symput('quantidade',aux);

run;

I don't know why but it doesn't work...

Regards,

Renata Matos