proc reg and almost zero rsquare

saslearner0903 — Sun, 16 Aug 2015 18:57:18 GMT

Hello~ guys~

I have a question about R-square using proc reg.

My sample consists of daily data.

I want to run regression by both group and year and to get R square by group.

Data structure is as follows.

group dept indpt1 indpt2 date year

1 19950101 1995

2 19960205 1996

3 19970310 1997

When I run the regression with following code, it generates residual=0 and Rsquare=1 for every observation.

However, it should not be the case since I got positive R-square with part of the sample using another program .

What is the problem of my code and how can I correct this problem?

Any comments are appreciated.

-------------------------------------------------------------------------------------------------------------------------------------------------------

proc sort data=examp; by group year;

run;

proc reg data = examp

noprint tableout edf usscp outest = result_examp;

by group year;

model dpt = indpt1 indpt2;

output out=result_R r=rsqr;

run;

Thank you.

Re: proc reg and almost zero rsquare

SteveDenham — Mon, 17 Aug 2015 13:05:49 GMT

Given your dataset, you are trying to regress a single point. You have only one record per group. I would suggest dropping the group variable from the by statement to get the relationship for each year across all groups, and then dropping year variable (and re-inserting group) to look at the relationship for each group across time. Does that make any sense relative to the objective of your analysis?

Steve Denham

topic proc reg and almost zero rsquare in SAS Procedures

proc reg and almost zero rsquare

Re: proc reg and almost zero rsquare