[Limdep Nlogit List] Fixed Effects Probit group size error

[William Greene]

Mr. Ohalloran.
Your pds indicator should give the number of periods,
not the index of the group. With your data as suggested
first below, I suggest you now use
REGRESS;LHS=one;RHS=one;Str=newins;Panel$
Then, after you do this, you can use
;PDS=_GROUPTI
in your probit commands. Your second data arrangement
is incorrect.
 One additional point, you cannot
include ONE in your SORT command; ONE is not in your data 
set.
/B. Greene


----- Original Message -----
From: "Patrick Ohalloran" <pohallor at monmouth.edu>
To: limdep at limdep.itls.usyd.edu.au
Sent: Monday, August 4, 2008 2:15:38 PM (GMT-0500) America/New_York
Subject: [Limdep Nlogit List] Fixed Effects Probit group size error

Hello,

I am attempting to fit a fixed effects probit model and I keep receiving the following message:

Error   678: group size for indiv.  1235=    31 is too large.

 

I've created the NEWIND as per LIMDEP manual section R6.5.3:

SORT;Lhs=IDCODE;Rhs=UNION,TENURE,EXPER,NONWHITE,FEMALE,AGE,ASVABSCR,EDUCATON,SMALFIRM,MEDFIRM,LGESTFRM,SHIFTWRK,OFFSHIFT,PROF,MANAGER,SALES,CLERIC,CRAFT,OPERAT,TRNSOPER,LABORER,SERVICE,PVTHOUS,CONSTRT,MFG,TRANPOR,WHOLESA,FINANCE,BUSINES,PERSERV,ENTERTN,PROFSRV,PUBADMN,HRSWORK,NUMJOBS,MARRIED,YOUNGKID,YR_1988,YR_1989,YR_1990,YR_1996,YR_1998,ONE$

CREATE; newind = Ndx(IDCODE,0)$

Consequently, my data is organized as follows:

IDCODE         YEAR  PRP     NEWIND

2                      1988    0          1

2                      1989    0          1

2                      1990    0          1

4                      1988    0          2

4                      1989    1          2

4                      1990    0          2

4                      1996    0          2

4                      1998    0          2

5                      1988    0          3

5                      1989    0          3

6                      1988    0          4

6                      1989    0          4

6                      1990    0          4

7                      1996    0          5

7                      1998    1          5

7                      2000    0          5

8                      1988    0          6

8                      1989    0          6

8                      1990    0          6

8                      1996    0          6

8                      1998    0          6

9                      1988    0          7

9                      1989    0          7

9                      1990    0          7

9                      1996    0          7

10                    1988    0          8

10                    1990    1          8

... all up to 

12,686             2000    1          8,551               

 

Here is the command line:

--> PROBIT;Lhs=PRP;Rhs=UNION,TENURE,EXPER,NONWHITE,AGE,ASVABSCR,EDUCATON,SMALFIRM,MEDFIRM,LGESTFRM,SHIFTWRK,OFFSHIFT,PROF,MANAGER,SALES,CLERIC,CRAFT,OPERAT,TRNSOPER,LABORER,SERVICE,PVTHOUS,CONSTRT,MFG

,TRANPOR,WHOLESA,FINANCE,BUSINES,PERSERV,ENTERTN,PROFSRV,PUBADMN,HRSWORK,NUMJOBS,MARRIED,YOUNGKID,YR_1988,YR_1989,YR_1990,YR_1996,YR_1998,ONE

;Pds=newind ;FixedEffects ;MARGIN$

 
If I'm understanding the LIMDEP manual correctly (see R6.5.2 - R6.5.3), I do have this set up correctly. 
 
It does run when I code this as:
--> CREATE

;newind2=NDX(IDCODE,1)$

Which results in:

IDCODE         YEAR  PRP     NEWIND2

2                      1988    0          1

2                      1989    0          2

2                      1990    0          3

4                      1988    0          1

4                      1989    1          2

4                      1990    0          3

4                      1996    0          4

4                      1998    0          5

5                      1988    0          1

5                      1989    0          2

6                      1988    0          1

6                      1989    0          2

6                      1990    0          3

7                      1996    0          1

7                      1998    1          2

7                      2000    0          3

8                      1988    0          1

8                      1989    0          2

8                      1990    0          3

8                      1996    0          4

8                      1998    0          5

...

 

Although is runs smoothly (no error messages) I don't think this is correct because it is grouping by year and not IDCODE.

Any assistance from those with more experience with fixed effects models?
 
Sincerely,
Patrick
 
 
Patrick L. O'Halloran
Assistant Professor
Department of Economics and Finance
School of Business Administration
Monmouth University
400 Cedar Avenue
West Long Branch, NJ 07764
pohallor at monmouth.edu
732.571.7530 - Telephone
732.263.5128 - Fax

_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au


-- 
Professor William Greene
Department of Economics
Stern School of Business
New York University
44 West 4th St., Rm. 7-78
New York, NY   10012
http://www.stern.nyu.edu/~wgreene
212.998.0876