 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
RE: st: ecm via xtrc
Thanks!
I am aware about you mean. Nevertheless, using panel macro data, one has to do with small N necessarily -- in particular small T. this involves serious ptoblem for testing order of integration (see for example levinlim Stata command). Nevertheless, those variable are often strongly autoregressive (estimating a simple ar(1)regression, the coeff for LDV is around 0.99). Then the risk of spurious regression is high. Consequently, one could be motivated to capture adjustment process via error correction model...
Moreover those data involve a causal heterogeneity modelling, so whay do not estimate an xtrc?
Accordingly, even if there are many stata command to tacking into account panel problems, such as xtwest for testing cointegration via ecm, I wonder whether I can't use xtrc in a ecm framework (see output below) because I have a small data set whether I need to looking for a different stata command?
In other words is it a data problem which we can't solve via any kind of models?
Best
Federico
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of
> Maarten buis
> Sent: Tuesday, May 13, 2008 6:00 PM
> To: [email protected]
> Subject: Re: st: ecm via xtrc
>
> One reason for differences in significance between your models has to
> do with the number of observations. In your "simple OLS procedure" you
> assume that you have 465 independent observations, which is way to
> optimistic. For example you can ask me on 10 different days whether or
> not I like chocolate, or you can ask 10 different people whether they
> like chocolate. In the latter case you have 10 indepedent pieces of
> information, but in the former case you don't, and you need to take
> that into account. I don't know whether -xtrc- is the best
> way of doing
> that, there are many other models in the -xt- family, see: -help xt-.
>
> -- Maarten
>
> --- "Podesta', Federico" <[email protected]> wrote:
>
> > Dear all,
> >
> > I am using a time series cross-section data set including 465
> > observations (31 annual observations for 15 countries). My dependent
> > variable is social security transfer (SSTRAN), while covariates are
> > left power (LEFTCUM), unemployment rate (UNEM), percentage of the
> > inactive population (DEPRATIO), and trade openness (TRADE).
> All these
> > variables seem non-stationary processes. Consequently, I mean to
> > estimate an panel error correction model- So if I estimate this kind
> > of model via a simple OLS procedure, the parameter for the lagged
> > dependent level variable which represents a measure of equilibrium
> > properties, is quite low (-0.014) (see belo the STATA output).
> > nevertheless, if estimate a random coefficient error
> correction model
> > via xtrc STATA command, the parameter for the lagged dependent level
> > variable increases strongly. In this case it is -0.30 (see STATA
> > output below). A part from the problem of the statistical
> > significance of the coefficient, this implies that the adjus!
> > tment process among variables is considerably faster.
> > On the basis of this, I wonder if it is statistically reasonable
> > estimate an error correction model using xtrc STATA command?
> > Why if one controls causal heterogeneity via a random coefficient
> > model, the adjustment process should be faster than a basic
> > specification?
> >
> > Thanks a lot in advance for any your help
> > Best regards
> > Federico Podest�
> >
> >
> > . reg dsstran lsstran dleftcum lleftcum dunem lunem ddepratio
> > ldepratio dtrade ltrade
> >
> > Source | SS df MS Number of obs
> > = 450
> > -------------+------------------------------ F( 9, 440)
> > = 28.44
> > Model | 73.1523504 9 8.12803894 Prob > F
> > = 0.0000
> > Residual | 125.736581 440 .285764957 R-squared
> > = 0.3678
> > -------------+------------------------------ Adj R-squared
> > = 0.3549
> > Total | 198.888931 449 .442959758 Root MSE
> > = .53457
> >
> >
> --------------------------------------------------------------
> ----------------
> > dsstran | Coef. Std. Err. t P>|t| [95% Conf.
> > Interval]
> >
> -------------+------------------------------------------------
> ----------------
> > lsstran | -.0138686 .0075559 -1.84 0.067
> -.0287187
> > .0009816
> > dleftcum | .0639292 .0739955 0.86 0.388
> -.0814992
> > .2093577
> > lleftcum | .0013023 .0037215 0.35 0.727
> -.0060118
> > .0086165
> > dunem | .4283545 .0308134 13.90 0.000
> .3677947
> > .4889143
> > lunem | -.0295548 .0094424 -3.13 0.002 -.0481127
> > -.0109969
> > ddepratio | .135711 .0872411 1.56 0.121
> -.0357501
> > .307172
> > ldepratio | .0020256 .0103268 0.20 0.845
> -.0182705
> > .0223216
> > dtrade | -.0188634 .0072521 -2.60 0.010 -.0331164
> > -.0046104
> > ltrade | .0024225 .0012992 1.86 0.063
> -.000131
> > .004976
> > _cons | .3227546 .3836955 0.84 0.401
> -.4313491
> > 1.076858
> >
> >
> >
> > . xtrc dsstran lsstran dleftcum lleftcum dunem lunem ddepratio
> > ldepratio dtrade ltrade
> >
> > Random-coefficients regression Number of
> obs =
> > 450
> > Group variable: cc Number of
> groups =
> > 15
> >
> > Obs per
> group: min =
> > 30
> >
> avg =
> > 30.0
> >
> max =
> > 30
> >
> > Wald
> chi2(9) =
> > 113.31
> > Prob > chi2
> =
> > 0.0000
> >
> >
> --------------------------------------------------------------
> ----------------
> > dsstran | Coef. Std. Err. z P>|z| [95% Conf.
> > Interval]
> >
> -------------+------------------------------------------------
> ----------------
> > lsstran | -.292154 .084294 -3.47 0.001 -.4573672
> > -.1269408
> > dleftcum | .1335818 .290957 0.46 0.646
> -.4366835
> > .703847
> > lleftcum | .1331348 .1545065 0.86 0.389
> -.1696924
> > .435962
> > dunem | .5189076 .1106382 4.69 0.000
> .3020606
> > .7357545
> > lunem | .0747792 .0588429 1.27 0.204
> -.0405507
> > .1901092
> > ddepratio | .0108599 .3924658 0.03 0.978
> -.758359
> > .7800788
> > ldepratio | -.0173162 .0737059 -0.23 0.814
> -.1617772
> > .1271448
> > dtrade | -.0033899 .0184207 -0.18 0.854
> -.0394938
> > .032714
> > ltrade | .0043399 .01749 0.25 0.804
> -.0299398
> > .0386197
> > _cons | 2.725573 2.747165 0.99 0.321
> -2.658771
> > 8.109917
> >
> --------------------------------------------------------------
> ----------------
> > Test of parameter constancy: chi2(140) = 353.60
> Prob > chi2
> > = 0.0000
> >
> > *
> > * For searches and help try:
> > * http://www.stata.com/support/faqs/res/findit.html
> > * http://www.stata.com/support/statalist/faq
> > * http://www.ats.ucla.edu/stat/stata/
> >
>
>
> -----------------------------------------
> Maarten L. Buis
> Department of Social Research Methodology
> Vrije Universiteit Amsterdam
> Boelelaan 1081
> 1081 HV Amsterdam
> The Netherlands
>
> visiting address:
> Buitenveldertselaan 3 (Metropolitan), room Z434
>
> +31 20 5986715
>
> http://home.fsw.vu.nl/m.buis/
> -----------------------------------------
>
>
> __________________________________________________________
> Sent from Yahoo! Mail.
> A Smarter Email http://uk.docs.yahoo.com/nowyoucan.html
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/