Reputation: 45
I'm trying to test different covariance structures inside macro with Proc Mixed.
%macro cov(type);
proc mixed data=tmp order=data;
class sub trt visit;
model var = trt visit trt*visit / S cl;
repeated visit /subject=sub type=&type.;
FitStatistics=min_var_&type.;
run;
%mend;
Some of the covariance structures I need to fit in model causes errors and I'm trying to find a way to execute this proc mixed statement only, if it doesn't cause error with value of &type.
I have been working with %sysfunc and but haven't been able to resolve this yet.
%IF %SYSFUNC(EXIST(min_var_&type.)) %THEN %DO;
data help_&type.;
set min_var_&type.;
run;
%end;
This produces these datasets correctly, but still log errors exists in log for those macro variables that can not be fitted.
Upvotes: 1
Views: 123
Reputation: 339
You can redirect the log to a file like that :
filename logfile "\\SERVER\LOG\mylog.log";
proc printto log=logfile new;
run;
And then when your PROC MIXED is finished, you can filter on the log file for the string "ERROR" :
....YOUR PROC MIXED...
/*come back to normal log*/
proc printto;
run;
/*check the log file*/
DATA CHECKLOG;
LENGTH ROWS $200;
LABEL ROWS = 'Messages from LOG';
INFILE "\\SERVER\LOG\mylog.log" TRUNCOVER;
INPUT ROWS &;
LINE = _N_;
IF SUBSTR(ROWS,1,5)='ERROR' /*OR SUBSTR(ROWS,1,7)='WARNING'*/ THEN
OUTPUT;
RUN;
You will have all the ERROR and (or WARNING if needed) in a dataset.
Then you have to check if the table is empty. If YES, you can continue your script.
You can do it via this method
proc sql;
select * from checklog;
run;
%put n=&sqlobs;
If sqlobs is greater than 0, then you have errors. You can check the sqlobs via a macro function like this :
%macro checklog;
proc sql;
select * from checklog;
run;
%if (&sqlobs>0) %then ...
%else ...
%mend checklog;
Upvotes: 1