Proc Rank

[Toni2]

i have 2 variables OURBAND and SURBAND and want use the proc rank 

Below are the proc freq of the variables and their ranks

I don't understand why for the variable OURBAND the proc rank creates only 1 rank (while has 5 different values) and at the same time the SURBAND with 9 different values has 5 ranks?

OURBAND
OURBAND	Frequency	Percent	Cumulative	Cumulative
			Frequency	Percent
1	20	0.67	20	0.67
2	92	3.07	112	3.73
3	42	1.4	154	5.13
4	7	0.23	161	5.37
5	2839	94.63	3000	100

SURBAND
SURBAND	Frequency	Percent	Cumulative	Cumulative
			Frequency	Percent
1	998	33.27	998	33.27
2	13	0.43	1011	33.7
3	32	1.07	1043	34.77
4	1102	36.73	2145	71.5
5	185	6.17	2330	77.67
6	248	8.27	2578	85.93
7	330	11	2908	96.93
8	37	1.23	2945	98.17
9	55	1.83	3000	100

Rank for Variable OURBAND
rank1	Frequency	Percent	Cumulative	Cumulative
			Frequency	Percent
0	3000	100	3000	100

Rank for Variable SURBAND
rank2	Frequency	Percent	Cumulative	Cumulative
			Frequency	Percent
0	998	33.27	998	33.27
3	1147	38.23	2145	71.5
7	433	14.43	2578	85.93
8	330	11	2908	96.93
9	92	3.07	3000	100

proc rank data=test_1 group=10 out=check_rank ties=low;
var &varnum;
ranks rank1-rank&vmcnt; 
run;

proc freq data=check_rank;
table ourband surband rank1 rank2;
run;

[PaigeMiller]

I don't know why PROC RANK is doing this either. You didn't show us the original data in data set TEST_1. A useful debugging technique is for you to look at data set TEST_1 with your own eyes, and see if you can figure it out.

If that doesn't help, then please show us (a portion of) the original data in data set TEST_1 as SAS data step code, which you can type in yourself, or by following these instructions.

@Toni2 don't make us ask to see your data. From now on, please just show it to us without us asking, in the form requested above. Thanks!

[Toni2]

thanks. I can't see anything strange in the TEST_1. TEST_1 contains RAW data

The data step which i used to create the TEST_1 is the below. i have not made any calculation

data test_1;
set input.raw_data;
keep &listvar;
run;

 Below is a small part of the TEST_1

SURBAND	OURBAND
9	3
6	5
4	5
4	5
9	5
4	5
4	5
7	5
5	5
7	5
7	5
6	3
4	5
6	5
6	2
5	5
3	3
7	5
7	5
7	5
6	5
7	5
5	5
9	3
5	5
5	5

[PaigeMiller]

PROC RANK is usually for continuous values. You have discrete values of OURBAND and SURBAND. It really doesn't make sense to try to rank these.

If you want to know if there are more 5s than 1s in a variable that takes discrete values, you would use PROC FREQ and not PROC RANK.

[Rick_SAS]

I think you need to get rid of the GROUPS=10 option. That option is telling the procedure to combine values that have different ranks into a single group, which is not what you want.

For more about the GROUP= option, see https://blogs.sas.com/content/sgf/2019/07/19/how-the-rank-procedure-calculates-ranks-with-groups-and... 

Look at the following example to see what the GROUPS=10 option is doing for your data:

data A;
input outband count;
do i = 1 to count;
Cnt + 1;
   output;
end;
drop i;
datalines;
1	20
2	92
3	42
4	7
5	2839
;

proc rank data=A out=rankLow ties=LOW;
   var outband;
   ranks rankLow;
run;
Title "Ranking";
proc sgplot data=rankLow;
scatter x=Cnt y=rankLow;
run;

proc rank data=A out=rankLowGroup ties=LOW GROUPS=10;
   var outband;
   ranks rankLow;
run;
Title "GROUPS=10 Option";
proc sgplot data=rankLowGroup;
scatter x=Cnt y=rankLow;
run;

[Toni2]

thanks for your response. I tried to run but get the below error (sorry i am not very experienced)

ERROR: Invalid DO loop control information, either the INITIAL or TO expression is missing or the BY expression is missing, zero, or invalid.

[Rick_SAS]

I assume you tried to cut and paste my example and accidentally did not copy all of the program. Try it again and be more careful.

[StatDave]

Intuitively speaking, OURBAND has only 161 observations of the 3000 that are less than 5. The 161-st value (in order) is only at the 5th percentile. If you run: proc univariate; var ourband; run; you will see that the 10-th percentile is already 5. So, all of those 161 observations will be in the first of the deciles you requested with GROUPS=10 (which is give rank value 0). Also, observations 162 through 300 (which have value 5) will also be in the first decile and since those values are tied with all of the remaining observations (also with value 5), then everything gets put in the first decile by GROUPS=10.