Addressing Heteroskedasticity and Autocorrelation in only one independent variable After running xtpmg Panel ARDL

[Ahmed Rashed]

Hi all. Please how to treat heteroskedasticity and Autocorrelation in GDP after running Panel ARDL pmg

After determining the optimal lags and performed the Hausman test., i run this syntax:

xtpmg d.FDIREI d.REO d.GDP, lr(l.FDIREI REO GDP) ec(ECT) replace pmg


Pooled Mean Group Regression
(Estimate results saved as pmg)

Panel Variable (i): ID Number of obs = 594
Time Variable (t): Year Number of groups = 33
Obs per group: min = 18
avg = 18.0
max = 18

Log Likelihood = -3281.471
------------------------------------------------------------------------------
D.FDIREI | Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
ECT |
REO | .0216125 .0037366 5.78 0.000 .0142889 .0289361
GDP | -.8550642 .2938029 -2.91 0.004 -1.430907 -.2792211
-------------+----------------------------------------------------------------
SR |
ECT | -1.071479 .147786 -7.25 0.000 -1.361134 -.7818239
|
REO |
D1. | .0413745 .3407124 0.12 0.903 -.6264095 .7091584
|
GDP |
D1. | -5.915913 5.516867 -1.07 0.284 -16.72877 4.896947
|
_cons | 15.5448 24.49003 0.63 0.526 -32.45479 63.54438

Then i run xthrtest:

. xthrtest Lag_FDIREI REO GDP, force

Heteroskedasticity-robust Born and Breitung (2016) HR-test on Lag_FDIREI REO GDP
Panelvar: ID
Timevar: Year
--------------------------------------------------------------------------------------+
Variable | HR-stat p-value | N maxT | balance? |
------------------------------+-----------------------+----------------+--------------|
Lag_FDIREI + 0.38 0.708 + 33 18 + balanced |
REO + -1.77 0.077 + 33 19 + balanced |
GDP + 2.31 0.021 + 33 19 + balanced |
--------------------------------------------------------------------------------------+
Notes: Under H0, HR ~ N(0,1)
H0: No first-order serial correlation.
Ha: Some first order serial correlation.

please how to treat heteroskedasticity and Autocorrelation in GDP after xtpmg panel ardl model

[George Ford]

cluster on the unit will resolve both

[Ahmed Rashed]

George Ford thank you for your response. I have already attempted to use the cluster ID option to address heteroskedasticity and autocorrelation. However, I found that the xtpmg command does not support the cluster option. Could you please advise on what I should do in this case?

I truly appreciate your valuable assistance.



. xtpmg d.FDI d.GREEN d.HD d.GDP d.Corr , lr(l.FDI GREEN HD GDP Corr ) ec(ECT) replace pmg cluster(id)

cluster option only meaningful with dfe
ignoring option and continuing...

Iteration 0: Log likelihood = -3213.7399 (not concave)
Iteration 1: Log likelihood = -3201.426
Iteration 2: Log likelihood = -3198.4952
Iteration 3: Log likelihood = -3196.5607
Iteration 4: Log likelihood = -3195.8093
Iteration 5: Log likelihood = -3195.7618
Iteration 6: Log likelihood = -3195.7614
Iteration 7: Log likelihood = -3195.7614

Pooled Mean Group Regression
(Estimate results saved as pmg)

Panel Variable (i): ID Number of obs = 594
Time Variable (t): Year Number of groups = 33
Obs per group: min = 18
avg = 18.0
max = 18

Log Likelihood = -3195.761
------------------------------------------------------------------------------
D.FDI | Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
ECT |
GREEN | .0086558 .0034439 2.51 0.012 .0019058 .0154058
HD | 2.203472 20.07417 0.11 0.913 -37.14118 41.54813
GDP | .0043967 .0019603 2.24 0.025 .0005546 .0082388
Corr | 5.413227 1.864287 2.90 0.004 1.759291 9.067162
-------------+----------------------------------------------------------------
SR |
ECT | -1.330891 .1767553 -7.53 0.000 -1.677325 -.9844564
|
GREEN |
D1. | .1567637 .429954 0.36 0.715 -.6859306 .999458
|
HD |
D1. | 6281.639 7512.059 0.84 0.403 -8441.727 21005.01
|
GDP |
D1. | -.1785376 .1256163 -1.42 0.155 -.4247411 .0676659
|
Corr |
D1. | 123.6281 200.3789 0.62 0.537 -269.1074 516.3636
|
_cons | 5.790713 22.08377 0.26 0.793 -37.49267 49.0741

[George Ford]

sorry. I missed the PMG option. I think the PMG option provides something akin to clustered errors by default (look at Pesaran, Shin and Smith 1999).

[Ahmed Rashed]

George Ford thank you for your response. I appreciate your valuable assistance.