Forum: help

Monitor Forum | Start New Thread Start New Thread
RE: Estimating Simultaneous Equation with Panel Data [ reply ]
By: Arne Henningsen on 2017-06-10 05:34
You can do the 'within' transformation manually (see equation in any textbook about panel data econometrics). I think that you can also use the 'plm' package for doing the 'within' transformation (see documentation of 'model.frame.pFormula').

RE: Estimating Simultaneous Equation with Panel Data [ reply ]
By: Gildardo Vargas on 2017-06-02 19:30
Dear Arne:

How can we perform the "within transformation" to the data set in order to apply systemfit? I know that there is a "within" argument in the plm package, but do not affect the data set.

Thank you in advance.

RE: Estimating Simultaneous Equation with Panel Data [ reply ]
By: Arne Henningsen on 2015-11-06 21:35
Dear Niklas

1. I just got some doubt about the suitability of the within transformation before doing a SUR estimation.

Assume a 2-equation SUR panel-data model with fixed individual effects:

y_1it = b10 + b11 * x_1it + d_1i + e_1it

y_2it = b20 + b21 * x_2it + d_2i + e_2it

with E[ e_1it e_2it ] != 0 (contemporaneously within-individual correlation) but E[ e_1it e_1jt ] = 0, E[ e_2it e_2jt ] = 0, E[ e_1it e_2jt ] = 0, and E[ e_2it e_1jt ] = 0 for i != j (no correlation between individuals), and E[ e_1it e_1js ] = 0, E[ e_2it e_2js ] = 0, E[ e_1it e_2js ] = 0, and E[ e_2it e_1js ] = 0 for t != s (no serial correlation).

Individual-specific mean values:

yBar_1i = b10 + b11 * xBar_1i + d_1i + eBar_1i

yBar_2i = b20 + b21 * xBar_2i + d_2i + eBar_2i


yTilde_1it = b11 * xTilde_1it + eTilde_1it

yTilde_2it = b21 * xTilde_2it + eTilde_2it


yTilde_1it = y_1it - yBar_1i

yTilde_2it = y_2it - yBar_2i

xTilde_1it = x_1it - xBar_1i

xTilde_2it = x_2it - xBar_2i

eTilde_1it = e_1it - eBar_1i

eTilde_2it = e_2it - eBar_2i

What happens with the estimation of the system of within-transformed equations if eBar_1i != 0 or eBar_2i != 0 for any i?

2. If a significant share of values of the dependent variables is censored, you need to use an estimator for censored dependent variables.

I also don't know any software package that implements the estimators discussed in Platoni, S., Sckokai, P., & Moro, D. (2012).

Best regards,

RE: Estimating Simultaneous Equation with Panel Data [ reply ]
By: Niklas Möhring on 2015-10-29 15:22
Dear Arne,

with respect to your answer below-
assuming the equation setting would be equal to the one suggested by Arunima:

1. Would the suggested approach to use within transformation still work, even though my panel is unbalanced?

2. Would the suggested "within transformation" approach still work if some of the independent variables are censored to zero for some individuals?

I have seen that such approaches are discussed, summarized and extended by Platoni, S., Sckokai, P., & Moro, D. (2012). Panel data estimation techniques and farm-level data models. American Journal of Agricultural Economics, aas072, but could not find that they have been implemented in software yet.

Thanks a lot for your help,


RE: Estimating Simultaneous Equation with Panel Data [ reply ]
By: Arne Henningsen on 2012-08-16 19:30
Dear Arunima

Given your model description, it seems to me that you have four different equations (with four different dependent variables: A,B,C,D). Hence, you do not have "the same (single) equation [...] for each individual". Thus, systemfit() can only perform a pooled regression with panel data (without manual interventions). However, if the number of individuals (and/or time periods) is not too large, you could add a dummy variable in each equation for each individual (and/or time period) (except for one individual/time period) in order to estimate a model with fixed individual (and/or time) effects. Alternatively, you could do a "within transformation" of your data set and then apply systemfit() to the "within-transformed" data in order to estimate a model with fixed individual and/or time effects.

I do not have sufficient information to answer your question regarding read.dta(). As this questions is not related to the systemfit package, I recommend that you ask it on the R-help mailing list (after reading the posting guide).

Best wishes,

Estimating Simultaneous Equation with Panel Data [ reply ]
By: Arunima Haldar on 2012-08-16 15:56
Dear Sir,

Can we estimate multiple Simultaneous Equation with Panel data using systemfit in R.The FAQ section "Function systemfit can estimate a system of equations with panel data if the same (single) equation is estimated for each individual. However, it cannot (yet) estimate arbitrary systems of multiple equations as panel data models (except for pooled regression)". Kindly clarify.

My prospective model:




Can such a model be estimated on R.Also I am unable to insert my data .dta file on R.I followed the following code.The software is unable to read my 3rd line of the code.

# input Stata file
mydata <- read.dta("c:/mydata.dta")


Thanks to:
Vienna University of Economics and Business University of Wisconsin - Madison Powered By FusionForge