How to use two-way fixed effect model when there are three identifying variable

[caveman529]

I have a panel dataset with three dimension: doctor, patient, and year.  I hope to include doctor and year fixed-effect in my regression.  I used

proc sort data=have; by doctor year; run;

proc panel data=HAVE;

     id doctor year;

     model y =  x1 x2 x3 x4 x5 x6 x7 / fixtwo;

run;

PROC TSCSREG DATA=HAVE;

   ID doctor year;

   MODEL y = x1 x2 x3 x4 x5 x6 x7 / fixtwo;

RUN;

But all I get is:

ERROR: Data set is not sorted in ascending sequence with

       respect to time series ID.  The current time period has year=2002 and

       the previous time period has year=2010 in cross section

       doctor=2.

How could I resolve this problem?  Thank you!

[DR_Majeti]

Hi,

Please check that error says dataset is not sorted as per variables in ID statement. .

Sort before this step.

[caveman529]

After sorting :

proc sort data=have; by doctor year; run;

it still reports errors...

[DR_Majeti]

Can you send me the sample data That i can try the output ..

[caveman529]

Hi, Majeti:

Thank you for your help on this!  I have attached the data file in zip. 

[DR_Majeti]

Hi ,

I have also got the same error, but I could rectify when only there are no duplicates for the year value. (But i know that it is wrong)

After doing that I got another error: when there is missing values for any observation the model we are applying is not accepting.

Please the check the combination of ID statement. Sorry, i don't have any idea with these kind of analysis.

[caveman529]

No problem.  Thanks!