Here is some sample made up data to show what I want.
data successes;
input code group something success;
datalines;
1104 1 8 0
1104 1 7 0
1105 1 2 0
1105 1 3 1
1105 1 4 0
1106 1 5 0
1106 1 6 1
1106 1 7 0
1106 1 8 0
1106 1 9 0
1106 1 1 0
1107 1 1 0
1107 1 3 0
1107 1 5 0
1107 1 7 0
1107 1 9 1
1108 1 2 0
1109 1 4 0
1109 1 6 0
1110 1 3 1
1110 1 2 1
1110 1 1 1
1111 2 2 0
1111 2 3 1
1111 2 4 0
1112 2 5 0
1112 2 6 1
1112 2 7 0
1112 2 8 0
1112 2 9 0
;
run;
For this data set, there are a number of observations for each code (in the real data anywhere between 1 and 25). A "1" in success means that that code was successful. In this data, codes 1104, 1108, and 1109 had no success in group 1. 1105, 1106, 1107, and 1110 had success.
What I want, is to adjust the data in such a way that I can do a Proc Freq to come up with a success percent by group. Group 1 has a success rate of 4/7 = 57%.
My thoughts were that I would make a successflag of some sort that was a 1 if any observation of a code had a success value of 1. Then de-duplicate based on code and successflag to create a data set I could run the FREQ against.
I tried
data mysuccesses;
mysuccessflag = 0;
if group = 1;
do until (last.code);
set successes;
by code;
if success = 1 then mysuccessflag = 1;
end;
run;
but that is wrong in many ways. Looping in the data step error. Help here appreciated.
I was trying to figure out how to loop through all the observations for a particular code and then set my flag to 1 if any observation had a success of 1.
I think I want a data set that looks like
data successes;
input code group something success mysuccessflag;
datalines;
1104 1 8 0 0
1104 1 7 0 0
1105 1 2 0 1
1105 1 3 1 1
1105 1 4 0 1
1106 1 5 0 1
1106 1 6 1 1
1106 1 7 0 1
1106 1 8 0 1
1106 1 9 0 1
1106 1 1 0 1
1107 1 1 0 1
1107 1 3 0 1
1107 1 5 0 1
1107 1 7 0 1
1107 1 9 1 1
1108 1 2 0 0
1109 1 4 0 0
1109 1 6 0 0
1110 1 3 1 1
1110 1 2 1 1
1110 1 1 1 1
1111 2 2 0 1
1111 2 3 1 1
1111 2 4 0 1
1112 2 5 0 1
1112 2 6 1 1
1112 2 7 0 1
1112 2 8 0 1
1112 2 9 0 1
;
run;
With that I could
proc sort data = successes out = cleanedsuccesses nodupkey;
by code;
run;
proc freq data = cleanedsuccesses;
where group = 1;
tables mysuccessflag;
run;
which gets me where I want to be.
But I'm open to suggestions. I can do this with SQL but I want to do it something like this.
Thank you.
... View more
