topic Re: Problem embedding macro variable string in output data set name in SAS Programming

Problem embedding macro variable string in output data set name

cau83 — Wed, 24 Jul 2013 13:33:55 GMT

I have a related post here with a separate question about a warning SAS gives: https://communities.sas.com/message/175096#175096

In the code below, on the line I've marked, there are two output data sets that are name using the values of 3 macro variables. I am looping through each forecasting series 48 times (6 different time periods on each of 8 different models).

%macro modeltest(Dept=,holdoutweeks=);
  %let modelname1=simple;
  %let modelname2=double;
  %let modelname3=linear;
  %let modelname4=damptrend;
  %let modelname5=addseasonal;
  %let modelname6=multseasonal;
  %let modelname7=winters;
  %let modelname8=addwinters;





  %do J = 1 %to 2; **loops through the 8 different model choices and does the 6 weeks of forecasting for each**;


       %let modelspec=&modelname&j;    (*********THE PROBLEM OCCURS HERE********)



       %DO I = 1 %TO 1;
            %let backpd=%eval((&holdoutweeks+1)*7 - (&i*7)); **backpd = # of days to go back but divided by i so it loops through one week at a time**;


            proc esm data=&Dept back=&BackPd lead=14 seasonality=7 nooutall outfor=fcast_&Dept._&i._&modelspec outstat=metrics_&Dept._&i._&modelspec out=_null_;   (*****THE PROBLEM IS HERE*****)
                 by queue_nm lob;
                 id date interval=day zeromiss=left;
                 forecast effort/ model=&modelspec;
            run;
       %end;
  %end;
%mend;



%modeltest(Dept=OSS,holdoutweeks=6);

The current specification works, producing the following in the log:

MPRINT(MODELTEST): proc esm data=OSS back=42 lead=14 seasonality=7 nooutall outfor=fcast_OSS_1_simple outstat=metrics_OSS_1_simple out=_null_;

However, I wanted to switch the order of &i and &modelspec so that the outfor dataset in this instance is "fcast_OSS_simple_1". But when I switch it around, using this (outfor=fcast_&Dept._&modelspec._&i), I get errors in the log that I'll selectively paste here:

WARNING: Apparent symbolic reference MODELNAME2_1 not resolved.

NOTE 138-205: Line generated by the macro variable "I".

fcast_OSS_&modelname2_1

ERROR 22-322: Syntax error, expecting one of the following: ;, (, BACK, DATA, LEAD, MAXERROR, NOOUTALL, OUT, OUTCOMPONENT, OUTEST,

OUTFOR, OUTPROCINFO, OUTSTAT, OUTSUM, PLOT, PRINT, PRINTDETAILS, SEASONALITY, SORTNAMES, STARTSUM.

WARNING: Apparent symbolic reference MODELNAME2_1 not resolved.

SYMBOLGEN: Macro variable DEPT resolves to OSS

SYMBOLGEN: Macro variable MODELSPEC resolves to &modelname2

SYMBOLGEN: Macro variable I resolves to 1

NOTE: Line generated by the macro variable "I".

fcast_OSS_&modelname2_1

ERROR 200-322: The symbol is not recognized and will be ignored.

Of course, I can just live with the alternate name (ultimately I'll be appending them all together anyway). I'd like an explanation for learning purposes-- a solution would be nice but not necessary.

Re: Problem embedding macro variable string in output data set name

AncaTilea — Wed, 24 Jul 2013 13:58:36 GMT

In your code would you please try the following:

(see my red markings)

%macro modeltest(Dept=,holdoutweeks=);

%let modelname1=simple;

%let modelname2=double;

%let modelname3=linear;

%let modelname4=damptrend;

%let modelname5=addseasonal;

%let modelname6=multseasonal;

%let modelname7=winters;

%let modelname8=addwinters;

%do J = 1 %to 2; **loops through the 8 different model choices and does the 6 weeks of forecasting for each**;

%let modelspec=&&modelname&j; (*********THE PROBLEM OCCURS HERE********)

%DO I = 1 %TO 1;

%let backpd=%eval((&holdoutweeks+1)*7 - (&i*7)); **backpd = # of days to go back but divided by i so it loops through one week at a time**;

proc esm data=&Dept back=&BackPd lead=14 seasonality=7 nooutall outfor=fcast_&Dept._&i._&modelspec outstat=metrics_&Dept._&i._&modelspec out=_null_; (*****THE PROBLEM IS HERE*****)

by queue_nm lob;

id date interval=day zeromiss=left;

forecast effort/ model=&modelspec;

run;

%end;

%mend;

%modeltest(Dept=OSS,holdoutweeks=6);

PS: has a nice, straightforward paper that explains the use of double ampersand.

Re: Problem embedding macro variable string in output data set name

cau83 — Wed, 24 Jul 2013 14:21:50 GMT

This was the answer to my other question (where I was receiving a warning w/o the &&), but it does not solve this issue.

I will check out the paper to which you linked.

Re: Problem embedding macro variable string in output data set name

cau83 — Wed, 24 Jul 2013 14:35:19 GMT

I ended up finding a more compact method for assigning the model name after I did this post (I noticed a separate thread here: https://communities.sas.com/ideas/1084). I now use this specification and without the &modelname reference the datasets no longer have an issue-- still do not understand why it was a problem in the first place.

%let models= simple/double/linear/damptrend/addseasonal/multseasonal/winters/addwinters;

%do j=1 %to 8;

%let modelspec=%scan(&models,&j);

outfor=fcast_&Dept._&modelspec._&i

Thanks.