Page 1 of 1

STATA and EVIEWS different results in a PMG panel model

Posted: Thu Jun 28, 2018 10:00 am
by gabyaaa999
I'm running this PMG panel model in STATA and EVIEWS:

ef = f(v1, p1, r1, gdp, gdp2)

The data is anual about ecological footprint (ef), GDP and its square, and institutional index like voice and accountabily (v1), political stability (p1) and regulatory quality (r1). All the variables are in logaritms execpt for the intitutional indexs. I have N=16 and T=13

I choose the automatic selection of lags with the Akaike criteria and with a max lags of 1 in dependent variable and regressors (becase it doesn't allows me to choose more). And I choose trend specification in constant (levels).

I did the same in STATA, but the estimated results are different. I think the problem is with the v1 variable because when I delete the v1 varible of the estimation, the results of both programs are exactly the same.

I checked the v1 variable in both date base (for eviews and for stata), because I thought it could be the problem, but it's the same.

Therefore, I don't know why could be happening.

Does anyone knows what could be happening?

Re: STATA and EVIEWS different results in a PMG panel model

Posted: Thu Jun 28, 2018 10:11 am
by EViews Gareth
How do the results differ?

Re: STATA and EVIEWS different results in a PMG panel model

Posted: Thu Jun 28, 2018 11:53 am
by gabyaaa999
EViews Gareth wrote:How do the results differ?


I'll show you the output:

This is the eviews output with V1 included:

Dependent Variable: D(EF)
Method: ARDL
Date: 06/28/18 Time: 13:31
Sample: 2003 2014
Included observations: 192
Maximum dependent lags: 1 (Automatic selection)
Model selection method: Akaike info criterion (AIC)
Dynamic regressors (1 lag, automatic): V1 P1 R1 GDP GDP2
Fixed regressors: C
Number of models evalulated: 1
Selected Model: ARDL(1, 1, 1, 1, 1, 1)
Note: final equation sample is larger than selection sample

Variable Coefficient Std. Error t-Statistic Prob.*

Long Run Equation

V1 -0.180772 0.045738 -3.952374 0.0002
P1 -0.142571 0.017121 -8.327453 0.0000
R1 -0.079239 0.045147 -1.755127 0.0826
gdp 7.727855 1.040159 7.429495 0.0000
gdp2 -0.402074 0.055583 -7.233734 0.0000


And this is the STATA output with V1 included:

xtpmg d(1).ef d(1).v1 d(1).p1 d(1).r1 d(1).gdp d(1).gdp2, lr(l.ef v1 p1 r1 gdp gdp2) ec(ec)

Panel Variable (i): p Number of obs = 192
Time Variable (t): t Number of groups = 16
Obs per group: min = 12
avg = 12.0
max = 12

Log Likelihood = 437.3097
------------------------------------------------------------------------------
D.lhe | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
ec |
v1 | .1266011 .0339684 3.73 0.000 .0600242 .193178
p1 | -.1197105 .0242471 -4.94 0.000 -.167234 -.072187
r1 | .2217932 .0508806 4.36 0.000 .122069 .3215175
gdp | -1.886396 .6508253 -2.90 0.004 -3.16199 -.610802
gdp2 | .1046843 .0379317 2.76 0.006 .0303395 .1790291



This is the eviews output with V1 excluded:

Dependent Variable: D(EF)
Method: ARDL
Date: 06/28/18 Time: 13:44
Sample: 2003 2014
Included observations: 192
Maximum dependent lags: 1 (Automatic selection)
Model selection method: Akaike info criterion (AIC)
Dynamic regressors (1 lag, automatic): P1 R1 GDP GDP2
Fixed regressors: C
Number of models evalulated: 1
Selected Model: ARDL(1, 1, 1, 1, 1)
Note: final equation sample is larger than selection sample

Variable Coefficient Std. Error t-Statistic Prob.*

Long Run Equation

P1 -0.094587 0.022594 -4.186395 0.0001
R1 -0.091164 0.039605 -2.301804 0.0233
GDP 4.815823 0.733492 6.565612 0.0000
GDP2 -0.242356 0.041302 -5.867926 0.0000


And this is the STATA output with V1 excluded:

xtpmg d(1).ef d(1).p1 d(1).r1 d(1).gdp d(1).gdp2, lr(l.ef p1 r1 gdp gdp2) ec(ec1)

Panel Variable (i): p Number of obs = 192
Time Variable (t): t Number of groups = 16
Obs per group: min = 12
avg = 12.0
max = 12

Log Likelihood = 421.2098
------------------------------------------------------------------------------
D.lhe | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
ec1 |
p1 | -.0945866 .0225938 -4.19 0.000 -.1388696 -.0503036
r1 | -.0911639 .0396054 -2.30 0.021 -.1687891 -.0135387
gdp | 4.815825 .733492 6.57 0.000 3.378208 6.253443
gdp2 | -.2423561 .0413018 -5.87 0.000 -.3233062 -.161406

As you can see, all the results in the output are different with v1, but without v1, they are exactly the same.

Thank you :D

Re: STATA and EVIEWS different results in a PMG panel model

Posted: Thu Jun 28, 2018 12:00 pm
by EViews Gareth
No idea - don't know enough about STATA's routine to comment, sorry.