Forum: open-discussion

Amoxicillin Out Of Date smasiajed [ Reply ] By: advittick mamePrereBm on 2020-06-30 18:21	[forum:47899]
For Sale Progesterone Internet Over Night Online https://cheapcialisll.com/ - Cialis Viagra 25 Mg 4 St <a href=https://cheapcialisll.com/#>Cheap Cialis</a> cialis half life

Description Of The Drug Fluconazole 100mg smasiajed [ Reply ] By: advittick mamePrereBm on 2020-06-26 03:32	[forum:47887]
Cialis Authentique Ligne https://bbuycialisss.com/ - buy cialis 5mg online Where To Buy Tretinoin Uk <a href=https://bbuycialisss.com/#>Cialis</a> Viagra For Woman That Lasts Up To 48hrs

Cialis Ou Generique smasiajed [ Reply ] By: advittick mamePrereBm on 2020-06-21 13:36	[forum:47877]
Can I Buy Alli Diet Pills Online https://bbuycialisss.com/ - Cialis Comprar Cialis Espana <a href=https://bbuycialisss.com/#>Buy Cialis</a> Cialis Viagra Meinung

RE: Warning Simplified Translog [ Reply ] By: João Silva on 2015-01-26 09:14	[forum:41843]
Dear Arne, That is clear to me. Thank you very much for the clarification and for your help on this matter. Regards, João

RE: Warning Simplified Translog [ Reply ]
By: Arne Henningsen on 2015-01-25 19:38

[forum:41837]

Dear João

I do not understand what you mean by "conservative". Anyway, I meant with my comment regarding the efficiency estimates the following: If you get an output-oriented technical efficiency estimate TE_it for individual (plot) i at time t, I would not too confident to claim that one could simply increase the yield by ( 100 * (1 - TE_it) / TE_it )% by just becoming technical efficient (i.e. without changing any input quantity), because the efficiency estimates often crucially depend on the model specification. For instance, if the estimated efficiency of plot A in year V is, say, TE_AV = 0.9 and the estimated efficiency of plot B in year W is, say, TE_BW = 0.8, it could easily be that you slightly change the model specification and get TE_AV = 0.95 and TE_BW = 0.88 and with another model specification you get TE_AV = 0.82 and TE_BW = 0.70. Hence, the estimated (absolute) levels of efficiency are often (not always) rather sensitive to changes of the model specification, while the (relative) efficiency rankings are often (not always) rather robust to changes of the model specification. I hope that my point is now better explained.

Best regards,
Arne

RE: Warning Simplified Translog [ Reply ]
By: João Silva on 2015-01-20 09:58

[forum:41813]

Dear Arne,

Thank you for your reply once more. I just did a lrtest comparing a simple Cobb-Douglas model and it seems my simplified Translog (the one with multicollinearity) 'performs better', i.e. as a higher log likelihood and the year effects are statistically significant:

Model 1: SFA_CD_TN (Cobb-Douglas model)
Model 2: SFA_TL_vu (Simplified Translog with Year)
#Df LogLik Df Chisq Pr(>Chisq)
1 13 -656.76
2 20 -645.00 7 23.52 0.001383 **

As for your question regarding the plots, you are right: I have a total of 1435 observations. My dataset contains 309 different (unique) plots but they are not all the same every year (unbalanced panel). Roughly there are 90 unique plots per year.

As for the TE scores, your comment is not completely clear to me. Do I understand well that the TE scores tend to be relatively conservative across different model specifications? Also, I do see that my mean TE score drops by about 10% if I include an inefficiency effects model (z variables).

Thank you for your attention,
João

RE: Warning Simplified Translog [ Reply ]
By: Arne Henningsen on 2015-01-19 20:20

[forum:41812]

Dear João

You can use a likelihood ratio test to test whether the interaction effects with YEAR have a statistically significant influence.

In your example, you use a kind of Cobb-Douglas functional form which is extremely restrictive. I suggest that you (at least) estimate a Translog specification and test whether the fit of the Translog model is significantly better than the fit of the Cobb-Douglas model.

It seems to me that you have 1435 observation rather than 1435 individual plots, right? How many different (individual) plots are included in your data set?

I would not pay too much attention to the estimated level of efficiency (84% on average in your case), because in my experience the estimated level of efficiency often depends on the model specification, e.g. the assumed functional form and the assumed distribution of the inefficiency term. However, the order of the efficiency scores is often rather robust against changes of the model specification.

Best regards,
Arne

RE: Warning Simplified Translog [ Reply ]
By: João Silva on 2015-01-19 09:53

[forum:41811]

Dear Arne,

Thanks a lot for your reply. The R^2 of the regression you mentioned is 1, so I think you are right regarding the multicollinearity problem between the YEAR variables. How could I deal with this? Just simply ignore the YEAR^2 and the YEAR interactions in the model?

I am working with an unbalanced panel with a total of 1435 plots across 17 periods of time (more or less 84 plots/year). Do you think this is an adequate sample size for the amount of variables I am including? I observe that my mean TE scores is quite high (84%) so I am not sure if I am using too many variables in the model.

Thank you,
Joao

RE: Warning Simplified Translog [ Reply ]
By: Arne Henningsen on 2015-01-17 21:30

[forum:41810]

Dear João

The covariance matrix could be singular, because there is too high correlation between the explanatory variables ("multicollinearity"), e.g. between the 8 explanatory variables that include the YEAR variable. What is the R^2 value of the following regression?

testReg <- lm( I(.5*YEAR^2) ~ LnNfert_ha + LnPfert_ha + LnKfert_ha + LnSeed_ha + LnInsAI_ha + LnHerbAI_ha + VARIETYd + SEASONd + YEAR + I(YEAR*LnNfert_ha) + I(YEAR*LnPfert_ha) + I(YEAR*LnKfert_ha) + I(YEAR*LnSeed_ha) + I(YEAR*LnInsAI_ha) + I(YEAR*LnHerbAI_ha), data=Panel_CLLS)

How many observations (plots, years) does your data set have?

Best regards,
Arne

Warning Simplified Translog [ Reply ]
By: João Silva on 2015-01-14 17:02

[forum:41787]

Dear all,

I want to estimate a simplified translog production frontier using the sfa() function of the frontier package. In my simplified translog I want to include as inputs the following variables: 6 production inputs (Nfert, Pfert, Kfert, Seed, HerbUse and InsUse), 2 dummy variables (Variety and Season), Year as variable (1979, 1980,...), YearSquared, and the interaction between Year and all 6 production inputs.

For this purpose I am running the following model in R:

TL <- sfa(LnYield_ha ~ LnNfert_ha + LnPfert_ha + LnKfert_ha + LnSeed_ha + LnInsAI_ha + LnHerbAI_ha + VARIETYd + SEASONd + YEAR + I(.5*YEAR^2) + I(YEAR*LnNfert_ha) + I(YEAR*LnPfert_ha) + I(YEAR*LnKfert_ha) + I(YEAR*LnSeed_ha) + I(YEAR*LnInsAI_ha) + I(YEAR*LnHerbAI_ha), data=Panel_CLLS, truncNorm=TRUE)

However, when I do run this code I obtain the following warning message:

In sfa(LnYield_ha ~ LnNfert_ha + LnPfert_ha + LnKfert_ha + LnSeed_ha + :
the covariance matrix of the maximum likelihood estimates is singular

I have played around with the model and realise that the problem comes from I(.5*YEAR^2) part. Could anyone explain me what this message exactly means and what are its implications?

Thank you for your attention.
Regards,
João