Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: mean group


From   Nahla Samargandi <[email protected]>
To   "[email protected]" <[email protected]>
Subject   RE: st: mean group
Date   Tue, 21 Jun 2011 21:00:37 +0100

Hi,
I would like to ask 
How I can impose the dummy variable in pmg estimator , 
if this is my pmg regressions
xtpmg d.gdpg d.lnhh d.lnca d.lnpop d.fl1, lr( l2.gdpg  fl1) pmg replace ec(ec) 
and 
I want to add
if dincome==0 
or
 if dincome==1
where I can add it in the formula above.

Bes regards,
N________________________________________
From: [email protected] [[email protected]] On Behalf Of Nahla Samargandi [[email protected]]
Sent: 21 June 2011 20:01
To: [email protected]
Subject: RE: st: mean group

Dear Markus,

Thank you indeed for your illistartion.

I understood why Stata dropped my lagged dependent variable,
MY data over the period 1980 to 2008
And you said that I need the level of GDP IN THE PREVIOUS PERIOD (t-1)
SO IF I want to have the initial gdp per capita in the dynamic panel data ,
can i  calculate it like this
1981-1980 , 1982-1981 , etc
And then I will get no value in 1980
It seems to me like taking the difference,

Regarding why I labelled the variables, because I tried to apply the mg using the same formula in your demo lab2, but unfortunately I could not.
  This the pmg result
xtpmg d.gdpg  d.lntradegdp  d.education d.lnca d.lnpop d.fl1, lr( l2.gdpg  fl1) pmg replace ec(ec)
Iteration 0:   log likelihood = -3332.5242
Iteration 1:   log likelihood = -3332.3657
Iteration 2:   log likelihood = -3332.3647
Iteration 3:   log likelihood = -3332.3647
Pooled Mean Group Regression
(Estimate results saved as pmg)
Panel Variable (i): cou                         Number of obs      =      1303
Time Variable (t): year                         Number of groups   =        52
                                                Obs per group: min =        12
                                                               avg =      25.1
                                                               max =        28
                                                Log Likelihood     = -3332.365
------------------------------------------------------------------------------
      D.gdpg |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
ec           |
         fl1 |    1.55204   .6458759     2.40   0.016     .2861466    2.817934
-------------+----------------------------------------------------------------
SR           |
          ec |  -.0517315   .0498468    -1.04   0.299    -.1494294    .0459665
  lntradegdp |  -.7894347   1.204742    -0.66   0.512    -3.150686    1.571817
             |
   education |
         D1. |   1.046573   7.152161     0.15   0.884    -12.97141    15.06455
             |
        lnca |
         D1. |   5.211071   2.081382     2.50   0.012     1.131637    9.290505
             |
       lnpop |
         D1. |  -124.6077   94.11905    -1.32   0.186    -309.0777    59.86222
             |
         fl1 |
         D1. |  -6.441647   1.344954    -4.79   0.000    -9.077709   -3.805585
             |
       _cons |   7.016672    6.00329     1.17   0.242     -4.74956     18.7829
------------------------------------------------------------------------------
xtpmg d.gdpg  d.lntradegdp  d.education d.lnca d.lnpop d.fl1, lr( l2.gdpg  fl1) pmg replace ec(ec)
now to order to estimate the mg
1-I generate
gen dgdp= d.gdpg (difference)
gen dfl1=d.fl1 (lag)
gen lgdpg= l. gdpg
nothing found where name expected
r(198);
Still I can’t apply mg using your command in my variables, I don’t know why, please could you help me.?
> 2- then
>  ECM - MG
> xtmg dgdp llY llW llK dlW dlK , trend robust
>
> 2- nlcom ((_b[llW])/(-_b[llY])) ((_b[llK])/(-_b[llY])) , post
>
b[llY])) , post

finally , how I cam impose the dummy variable in the pmg and mg , because when I typed
xtpmg d.gdpg  d.lntradegdp  d.education d.lnca d.lnpop d.fl1, lr( l2.gdpg  fl1) pmg replace ec(ec) if  dincome= =1
It says   option if not allowed r(198);

Best regards,
N
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

From: [email protected] [[email protected]] On Behalf Of Markus Eberhardt [[email protected]]
Sent: 21 June 2011 15:04
To: [email protected]
Subject: Re: st: mean group

Well, Stata dropped your lagged dependent variable because it's not a
lagged dependent variable in the panel sense! "lngdpc0", I would
guess, is log of GDP per capita IN THE BASE YEAR. You cannot have a
variable that does not change over time in these dynamic regressions.
This is different from a cross-section regression, like in Mankiw,
Romer & Weil (1992, QJE). You need to have a variable which represents
the level of GDP IN THE PREVIOUS PERIOD (t-1), not in the base year.
Also, why do you label all variables with the prefix d-, even if they
seem to be levels variables?
In the example you quote, which is from my demo ado, you can see that
there are growth rates (d-  prefix) and lagged levels (l- prefix) to
clearly distinguish them from each other. If you use this setup you'll
be able to run an ECM which delivers the results you're looking for.
m

Markus Eberhardt
ESRC Post-doctoral Research Fellow, Centre for the Study of African
Economies, Department of Economics, University of Oxford
Stipendiary Lecturer, St Catherine's College, Oxford

web: http://sites.google.com/site/medevecon/home
email: [email protected]
twitter: http://twitter.com/sjoh2052
mail: Centre for the Study of African Economies, Department of
Economics, Manor Rd, Oxford OX1 3UQ, England




On 21 June 2011 14:56, Nahla Samargandi <[email protected]> wrote:
> Thank you indeed for replying.
> There are more that one command regarding the MG which of them that use to estimate Error correction Model  - TO estimate a long - short Run relationship  (ARDL ) Autoregressive distributed lag )
>
> 1-
>  ECM - MG
> xtmg dlY llY llW llK dlW dlK , trend robust
>
> 2-
> nlcom ((_b[llW])/(-_b[llY])) ((_b[llK])/(-_b[llY])) , post
>
> I got this result from the first one,
> gen dgdp=l.gdp etc
>
>  xtmg  dgdpg lngdpc0   deducation  dlnca  dlnpop  dfl1, trend
>
> Pesaran & Smith (1995) Mean Group estimator
> All coefficients present represent averages across groups (cou)
> Coefficient averages computed as unweighted means
> Mean Group type estimation                      Number of obs      =      1306
> Group variable: cou                             Number of groups   =        52
>                                                Obs per group: min =        12
>                                                               avg =      25.1
>                                                               max =        28
>                                                Wald chi2(4)       =     36.51
>                                                Prob > chi2        =    0.0000
> ------------------------------------------------------------------------------
>       dgdpg |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>     lngdpc0 |  (omitted)
>  deducation |    8.75228   11.61385     0.75   0.451    -14.01045    31.51501
>       dlnca |   5.732087   1.806081     3.17   0.002     2.192234     9.27194
>      dlnpop |   500.2783   321.8142     1.55   0.120    -130.4659    1131.022
>        dfl1 |  -6.580965   1.245776    -5.28   0.000    -9.022641    -4.13929
>  __000007_t |   .2274885    .151299     1.50   0.133    -.0690521    .5240291
>       _cons |  -11.20187   8.003771    -1.40   0.162    -26.88897    4.485233
> ------------------------------------------------------------------------------
> Root Mean Squared Error (sigma): 4.1932
> Variable __000007_t refers to a group-specific linear trend.
> Share of group-specific trends significant at 5% level: 0.038 (= 2 trends)
>
> however , I cant interpret the result, because I need to know the short and long run effect , I don't think the estimation above is the right one for this.
>
> regarding this command
>
> nlcom ((_b[llW])/(-_b[llY])) ((_b[llK])/(-_b[llY])) , post
> could you tell me is this the right one ? if yes, where i have to replace my variable?
>
> Best regards,
> Nahla
>
> ________________________________________
> From: [email protected] [[email protected]] On Behalf Of Markus Eberhardt [[email protected]]
> Sent: 21 June 2011 14:20
> To: [email protected]
> Subject: Re: st: mean group
>
> xtpmg (Blackburn and Frank) with option -mg- gives you an ECM version
> (well, it reports the implied long-run results as well as the
> short-run results; an option to see the results country by country is
> also given).
>
> ARDL means autoregressive distributed lag, which is a levels
> regression with lagged dependent variable and contemporaneous and
> lagged covariates. This is mathematically equivalent to an ECM, which
> has the first difference of y as dependent variable and then adds
> lagged levels of y and x as well as contemporaneous and lagged
> differences of x (and lagged ones of y, too) as covariates. Hendry
> (1995) 'Dynamic Econometrics' has a discussion for a single time
> series how ARDL is the encompassing specification for a lot of dynamic
> models, including ECM. In terms of interpretation of coefficients you
> will get the same long-run results, although they're constructed
> differently. Furthermore, an ECM approach allows you to impose a
> long-run relationship (y-beta*x) as 'ecm' variable in order to focus
> on short-run dynamics and the speed of convergence/error correction
> mechanism.
>
> If you create lags manually (gen lx=l.x etc.) you can also use my xtmg command.
>
> Both xtpmg and xtmg commands can be found using -findit- in Stata.
>
> Best
> m
>
> Markus Eberhardt
> ESRC Post-doctoral Research Fellow, Centre for the Study of African
> Economies, Department of Economics, University of Oxford
> Stipendiary Lecturer, St Catherine's College, Oxford
>
> web: http://sites.google.com/site/medevecon/home
> email: [email protected]
> twitter: http://twitter.com/sjoh2052
> mail: Centre for the Study of African Economies, Department of
> Economics, Manor Rd, Oxford OX1 3UQ, England
>
>
>
>
> On 21 June 2011 13:30, Nahla Samargandi <[email protected]> wrote:
>> Hi,
>> I would like know the right command for mean group MG  estimator in error correction model form (ARDL)developed by Pesaran 1999,
>>
>> thank you in advance for your help
>> *
>> *   For searches and help try:
>> *   http://www.stata.com/help.cgi?search
>> *   http://www.stata.com/support/statalist/faq
>> *   http://www.ats.ucla.edu/stat/stata/
>>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index