Forum: help

RE: Fixed Effect Stochastic Frontier Model using Panel Data setting [ Reply ]
By: Arne Henningsen on 2017-02-10 22:41

Dear Luca

There is no SFA model that people usually call "fixed effects" model.

Your model "I" assumes that the same 'frontier' applies to all farmers and time periods, inefficiency is time-invariant, and noise is time-variant and (even iid). Hence, time-invariant individual effects are absorbed by the inefficiency term (to the extent that the distributional assumption of the inefficiency term allows for this).

Your model "II" assumes that the frontier varies between farmers due to time-invariant (fixed) effects ofn the frontier and inefficiency and noise are time-variant (and even iid). Hence, time-invariant individual effects are absorbed by the 'fixed effects' (coefficients of the variable that indicate the individual farmer).

If you think that time-invariant effects only indicate differences in inefficiency (e.g. differences in managerial ability) but do not indicate differences in the 'frontier', you should estimate model "I". If you think that time-invariant effects only indicate differences in the frontier (e.g. differences in soil quality) but doe not indicate differences in inefficiency, you should estimate model "II".

Please note that a "Warning message" gives a warning and does *not* indicate an error.

Best,
Arne

RE: Fixed Effect Stochastic Frontier Model using Panel Data setting [ Reply ]
By: Luca Elisei on 2017-02-08 17:16

[forum:43853]

Dear Arne,

I need to clarify, how estimate a fixed effect model by sfa.

I)
# Error Components Frontier (Battese & Coelli 1992)
# with time-invariant efficiencies
riceTimeInv <- sfa( log( PROD ) ~ log( AREA ) + log( LABOR ) + log( NPK ), data = ricePanel )
summary( riceTimeInv )

Is a Fixed-effect model?
Can this control unobserved time-invariant factors?

II)
# Error Components Frontier (Battese & Coelli 1992)
# with "true" fixed individual effects and observation-specific efficiencies
riceTrue <- sfa( log( PROD ) ~ log( AREA ) + log( LABOR ) + log( NPK ) +
factor( FMERCODE ), data = riceProdPhil )
summary( riceTrue )

Is a Fixed-effect model?

the model produces this error:
Warning message:
In sfa(log(PROD) ~ log(AREA) + log(LABOR) + log(NPK) + factor(FMERCODE), :
the parameter 'gamma' is close to the boundary of the parameter space [0,1]: this can cause convergence problems and can negatively affect the validity and reliability of statistical tests and might be caused by model misspecification

I tried in my analysis and produces the same error. How can I resolve it?

Thanks,
BR,

Luca

RE: Fixed Effect Stochastic Frontier Model using Panel Data setting [ Reply ] By: Dhanasekaran Kuppusamy on 2015-12-07 14:17	[forum:42776]
Dear Arne, I used another track. Thanks a lot. Best regards Dhanasekaran

RE: Fixed Effect Stochastic Frontier Model using Panel Data setting [ Reply ] By: Arne Henningsen on 2015-12-07 07:22	[forum:42773]
Dear Dhanasekaran Kuppusamy Which fixed-effects stochastic frontier model do you mean? The one with individual-specific fixed (time-invariant) efficiency scores, Bill Greene's "true fixed effects" model, or something else? Best regards, Arne

Fixed Effect Stochastic Frontier Model using Panel Data setting [ Reply ] By: Dhanasekaran Kuppusamy on 2015-12-06 16:58	[forum:42768]
Dear A. H, How can we estimate the Fixed Effect Stochastic Frontier Model in Panel Data setting?. What is the R code for the data [data = riceProdPhil] in"'frontier"? kindly let me knowfor understanding the model. Best regards K.Dhanasekaran