topic Re: Issues with PROC RANK in SAS Procedures

Issues with PROC RANK

Doug____ — Fri, 25 Jan 2019 18:03:16 GMT

I am using PROC RANK with by group processing (pre-sorted). Even though the values are different, PROC RANK returns all 1s for the rank values. Many values are being ranked by the same by group processing. It does not happen with all variables, but it does occur with several of them.

This is the list of example values for a parameter within a single by group (all are rank = 1 in the output dataset):

6.5

3.4

0.77777

0.33333

0.88888

0.666666

Re: Issues with PROC RANK

PaigeMiller — Fri, 25 Jan 2019 18:07:08 GMT

Show us your code. Click on the running man icon and paste the code into the window that appears (DO NOT SKIP THIS STEP)

Re: Issues with PROC RANK

Reeza — Fri, 25 Jan 2019 18:11:36 GMT

And the log please.

Re: Issues with PROC RANK

PeterClemmensen — Fri, 25 Jan 2019 18:37:17 GMT

Show us your code please.

Re: Issues with PROC RANK

Reeza — Fri, 25 Jan 2019 18:40:30 GMT

I cannot replicate your issue.

data have;
input X;
cards;
6.5
0
3.4
0.77777
0.33333
1
2
0
0
0.88888
2
0
0.666666
;;;;

proc rank data=have out=want groups=4;
var x;
Ranks x_rank;
run;

proc freq data=want;
table x_rank;
run;

Rank for Variable X
x_rank	Frequency	Percent	Cumulative Frequency	Cumulative Percent
0	4	30.77	4	30.77
1	2	15.38	6	46.15
2	3	23.08	9	69.23
3	4	30.77	13	100.00

@Doug____ wrote:

I am using PROC RANK with by group processing (pre-sorted). Even though the values are different, PROC RANK returns all 1s for the rank values. Many values are being ranked by the same by group processing. It does not happen with all variables, but it does occur with several of them.

This is the list of example values for a parameter within a single by group (all are rank = 1 in the output dataset):

6.5

0

3.4

0.77777

0.33333

1

2

0

0

0.88888

2

0

0.666666

Re: Issues with PROC RANK

Doug____ — Fri, 25 Jan 2019 19:12:56 GMT

proc rank data = CM_BY_FA_S out = CM_BY_FA_ORG_RANKS descending ties = low;

by Manager;

VAR

Var1

Var2

Var3, etc.;

Ranks

Var1_R

Var2_R

Var3_R, etc.;

run;

NOTE: There were 48 observations read from the dataset WORK.CM_BY_FA_S.

NOTE: The data set WORK.CM_BY_FA_ORG_RANKS has 48 observations and 280 variables.

NOTE: PROCEDURE RANK used (Total process time):

real time 0.15 seconds

cpu time 0.07 seconds

There are 140 variables to rank in the dataset.

Re: Issues with PROC RANK

Reeza — Fri, 25 Jan 2019 19:14:51 GMT

Still don't get all 1s.

You can use this macro if you want to more easily get the ranks out:

https://gist.github.com/statgeek/7cffd06ebc3bc9c78b4f6a5b4538b053

Re: Issues with PROC RANK

Rick_SAS — Fri, 25 Jan 2019 19:41:09 GMT

The code you posted uses TIES=LOW. If you are getting all ranks equal to 1, that means that all values for a certain variable are the same. See the following example. The X variable is a constant column, so the ranks are always 1.

data Have;
set Sashelp.Class;
X = 99;
run;

proc rank data = Have out = CM_BY_FA_ORG_RANKS descending ties = low;
VAR   Height   Weight   Age   X;
Ranks Height_R Weight_R Age_R X_R;
run;

proc print;
run;