[Limdep Nlogit List] (no subject)

[O'Halloran, Patrick]

Hello,

I am attempting to fit a fixed effects ordered probit model and continually receive the following error message:  

Error   678: group size for indiv.  1008=    20 is too large.

 

Here is a log of the commands I am submitting and the output I receive.

 

--> ORDERED; Lhs=SATIS;

Rhs=ONE,PRP,MARRIED,SUPERVIS,TENURE,UNION,FIRMSIZE,PENSION ,HOURS,CHLDCARE,PAIDVAC,HEALTH,PUBLIC,PROMO,INJURY,EDUCAT,AFQT,FEMALE,BLACK

,HISP,SMSA,URBAN,NORTHEST,NORTHCEN,WEST,PROF,MANAGER,SALES,CLERIC,CRAFT

,OPERAT,TRNSOPER,LABORER,FARM,FRMLABR,SERVICE,PVTHOUS,AG,CONSTRT,MFG

,TRANPOR,WHOLESA,FINANCE,BUSINES,PERSERV,ENTERTN,PROFSRV,PUBADMN,LNWAGE

;$

Normal exit from iterations. Exit status=0.

 

OUTPUT OMMITED

 

--> ORDERED;Lhs=SATIS;

Rhs=PRP,MARRIED,SUPERVIS,TENURE,UNION,FIRMSIZE,PENSION

,HOURS,CHLDCARE,PAIDVAC,HEALTH,PUBLIC,PROMO,INJURY,EDUCAT,SMSA,URBAN,NOR...

,NORTHCEN,WEST,PROF,MANAGER,SALES,CLERIC,CRAFT

,OPERAT,TRNSOPER,LABORER,FARM,FRMLABR,SERVICE,PVTHOUS,AG,CONSTRT,MFG

,TRANPOR,WHOLESA,FINANCE,BUSINES,PERSERV,ENTERTN,PROFSRV,PUBADMN,LNWAGE

;PDS=id

;TIME = YEAR

;FIXED EFFECTS

;MARGIN$

  Error   678: group size for indiv.  1008=    20 is too large.

 

I have an unbalanced panel of 9,563 observations among 4,133 individuals.  The variable id uniquely identifies each individual i = 1 ... 4,133 and the t = 1996, 1998, 2000. The regressors are a combination of time variant and time invariant observations (i.e., gender does not change over time, but one's occupation may change from period to period.) 

Here is the pattern of observations:

 

    Freq.  Percent    Cum. |  Pattern

 ---------------------------+---------

     2139     51.75   51.75 |  111

      557     13.48   65.23 |  11.

      410      9.92   75.15 |  1..

      315      7.62   82.77 |  .11

      280      6.77   89.55 |  1.1

      229      5.54   95.09 |  .1.

      203      4.91  100.00 |  ..1

 ---------------------------+---------

     433    100.00         |  XXX

 

Thanks in advance for any assistance,

Patrick