Panel ARDL Model: Detecting Heteroskedasticity and Autocorrelation

[Ahmed Rashed]

Hi all,

This is my first post on Statalist. Thank you in advance for your help and guidance!

After completing the data cleaning stage, I detected cross-sectional dependence in the data. I found that the data employed exhibits cross-sectional dependence. Consequently, I used second-generation unit root tests and obtained mixed results:

FDI (my dependent variable) and GDP (one of my independent variables) are stationary, I(0),
EO (electricity output), the second independent variable, is non-stationary.
The aim of my study is to estimate the short- and long-run effects on the dependent variable (FDI) in 33 countries over 19 years. Based on the literature, I determined that applying a panel ARDL model is appropriate for my case. I followed these steps:

To determine the optimal lags:


forval i = 1/33 {
ardl FDIREI EO GDP if (ID == `i'), maxlag(2 2 2)
matrix list e(lags)
di
}
Based on the lag outcomes, the most common lags for each variable across countries are:

One lag for FDIREI
Zero lags for both EO and GDP
To determine which estimator to use:

xtpmg d.FDIREI d.EO d.GDP, lr(l.FDIREI l.EO l.GDP) ec(ECT) replace dfe
xtpmg d.FDIREI d.EO d.GDP, lr(l.FDIREI l.EO l.GDP) ec(ECT) replace pmg
hausman dfe pmg, sigmamore

Then I run the ARDL PMG panel data model:

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

My questions:

Are these steps correct so far?
What commands should I use to detect heteroskedasticity and autocorrelation?
How can I address these issues if they are present?
Thank you for your time and support!

Best regards,
Rashed

[George Ford]

cluster on the unit. done.