Proc Panel Error: No residual in the model to analyze. Computations will not be performed.

[applemonster]

I run the following code:

proc panel data=rrd_come.citydata_4;
id City TimeID;
model LoanNum = NewConfirmed NewSuspected / fixone;
run;

and I got such error:

ERROR: No residual in the model to analyze. Computations will not be performed.
NOTE: The transformed regression does not have full rank. Be aware of possible multicollinearity
      and/or identification problems before using the FixOne method results.
NOTE: PROCEDURE PANEL used (Total process time):
      real time           4.64 seconds
      cpu time            4.68 seconds

Can any generous friends help me out? I can send the data through email if necessary.

Thanks in advance!

[SteveDenham]

Here is the best clue as to what is happening:

NOTE: The transformed regression does not have full rank. Be aware of possible multicollinearity
      and/or identification problems before using the FixOne method results.

I suspect that one or the other of NewConfirmed and NewSuspected is exactly collinear by parts with the other, so that including both leads to the non-full rank issue.

Have you run this with each of these variables alone, and maybe looked at the solutions?  I suspect they may be the same, to within roundoff error.

SteveDenham

[applemonster]

Multicollinearity usually happens when variables describe multiple sides of the same thing. However, this may be not my case because NewConfirmed here is the daily increase of new confirmed COVID19 cases and NewSuspected here is the daily increase of new suspected COVID19 cases. So, theoretically, there should be no multicollinearity problem.

[SteveDenham]

Well, that leaves identification problems. In this case, that sounds like either of the independent variables gives exactly the same results per City and TimeID.  What happens if you drop the fixone option?

SteveDenham