Dear Gretl team. First of all, Happy New Year to the whole list.

Recently reviewing the Wooldridge introductory manual Introductory Econometrics: A Modern Approach (7 edition) I have seen a thing that has caught my attention and perhaps could be of interest for the list or an option as a future feature to add to Gretl. In Chapter 14 on Panel Data, section 14.3 Correlated Random Effects (CRE) Wooldridge explains what to me is the Mundlak's (1974) approach to estimate fixed effect models (i.e to expand the linear panel data model by adding the between transformation of time-varying variables into the model). The purpose of Wooldridge is to highlight that coefficients CRE are the same that coefficients in the fixed in their words "adding the time average x and using random effects is the same as subtracting the time averages and using pooled OLS". One of the benefits of this approach is that as known, by using fixed effects (within estimator) it is not possible to obtain coefficients on time constant variables, but the CRE estimator approach rectifies that.

On Tue, 11 Jan 2022, JOSE FRANCISCO PERLES RIBES wrote: > Dear Gretl team. > > First of all, Happy New Year to the whole list. > > Recently reviewing the Wooldridge introductory manual Introductory > Econometrics: A Modern Approach (7 edition) I have seen a thing that has > caught my attention and perhaps could be of interest for the list or an > option as a future feature to add to Gretl. > > In Chapter 14 on Panel Data, section 14.3 Correlated Random Effects (CRE) > Wooldridge explains what to me is the Mundlak's (1974) approach to estimate > fixed effect models (i.e to expand the linear panel data model by adding > the between transformation of time-varying variables into the model). The > purpose of Wooldridge is to highlight that coefficients CRE are the same > that coefficients in the fixed in their words "adding the time average x > and using random effects is the same as subtracting the time averages and > using pooled OLS". > > One of the benefits of this approach is that as known, by using fixed > effects (within estimator) it is not possible to obtain coefficients on > time constant variables, but the CRE estimator approach rectifies that. > > Panel command in Gretl performs the wihin, GLS and between estimators (most > used in practice). Although the approach described would not be very > difficult to implement using the standard panel tools of Gretl, perhaps it > would be a nice possibility to implement a command or an option in the > panel command to perform this approach. Thanks, José, this is a nice suggestion. If I'm not mistaken, you're suggesting to have an option to the "panel" command ("--mundlak", maybe?) that does more or less the following: <hansl> open abdata.gdt Ind7 = IND == 7 # a time-invariant dummy, just for the example list X = w k ys Ind7 list Mundlak = null loop foreach i X if !isconst($i, 0) series ta_$i = pmean($i) Mundlak += ta_$i endif endloop panel n const X Mundlak --random ### check that \hat{\beta} is in fact what you get from FE panel n const X Mundlak --fixed-effects </hansl> Is my understanding correct? ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------_______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/

Am 11.01.2022 um 15:40 schrieb JOSE FRANCISCO PERLES RIBES: > Dear Riccardo: > suggesting to have an option to the "panel" command ("--mundlak", maybe?) > Yes indeed, it is my suggestion. > > This looks like a natural target for a function package, no? (I mean, it doesn't have to be natively coded in C.) cheers sven _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/