sas >> rank regression

by Dana201052 » Thu, 29 May 2008 05:42:19 GMT

Hello All,

I've been stuck for a couple of days on running a rank regression of
some sort. I've gotten the code to run but with nothing remotely close
to what I need.

What I'm trying to do in SAS is to run a multiple regression analysis
by quintiles and deciles. I want to rank one of the variable into
quintiles and then deciles, and then run a regression for each level.
I want to determine whether the levels of the ranked variable matter.

Any assistance would be greatly appreciated.


sas >> rank regression

by Dana201052 » Thu, 29 May 2008 13:26:45 GMT

I just answered my own question so no response is necessary. This is
what I came up with that seems to work.

proc rank data=final3 out=rank1 ties=low groups=5;
var acc;
ranks accrank;
proc print data=rank1;
Title 'Rank by Quintile';
run; quit;


Similar Threads

1. Proc Rank: Filling in missing ranks

2. proc rank, ranks var assigment

data xx;
do i=1 to 49;

proc rank data=xx out=tert group=3;
var x;
ranks tert;

proc freq;
table tert;

                                  Cumulative    Cumulative
 tert    Frequency     Percent     Frequency      Percent
    0          16       32.65            16        32.65
    1          17       34.69            33        67.35
    2          16       32.65            49       100.00

I would like to get tertile assignment as :

0 - 17
1 - 16
2 - 16

i.e., low tertile has the high priority to get filled. It seems
that proc rank fills in the mid tertile first.

Any idea?


Ya Huang

3. Deming regression, orthogonal regression

4. Availability of Friedman Rank Sum Test

Hi SAS gurus-


Is anyone aware of the availability of the nonparametric Friedman rank
sum test in the SAS documentation.?


It is equivalent  to two factor  ANOVA without replication,  using


Thanks in advance.

Joe Earley

5. TTEST OR wilcoxon signed rank

6. Simple question about locking variable class in Wilcoxon rank

7. Rank order of records in an ordinal logistic

8. Wilcoxon signed rank test

The null hypothesis tested in wilcoxon signed rank test is that mean or median is equal to zero. My question is: how can I define de null hypothesis? In proc univariate the null hypothesis default of wilcoxon rank test is the mean equal to zero or it is the median equal to zero?
Thank you
Best Regards