[Sebastian Kripfganz]

There is nothing that prevents you from conducting unit-root tests. Note that commands for conventional unit-root tests usually do not carry out tests for seasonal unit roots or tests against explosive alternatives. It is often sufficient to rule those out based on our knowledge about the time series in question - e.g., explosive time series are quite rare in practice.

[ahmed mohamed ak]

"Decide about the candidate variables xt that are assumed to be long-run forcing
for yt. These variables can be either I(0) or I(1). No pretesting is necessary
unless we suspect that a variable might be I(2). Stationary variables zt that are
suspected to affect the short-run dynamics—but not the long-run equilibrium—
can be added to the ARDL model as well. If there is doubt about the (trend)
stationarity of zt, unit-root tests can be carried out."
In this paragraph, you did not mention the degree of integration of the dependent variable. What is the degree of integration of the dependent variable?

[Sebastian Kripfganz]

The dependent variable can be I(0) or I(1). Again, no pretesting is necessary.

[ahmed mohamed ak]

What is your opinion, Professor, on this table?

[Simra Temmy]

Hi,
In my current study, if I don’t use nocons in the ARDL code, I encounter a heterogeneity and serial correlation problem. However, when I do use nocons, the heterogeneity problem disappears, even though some variables become insignificant.
So, the questions is what happens if I use the code with nocons? How can I explain why do I use nocons?
Thank you...

[Sebastian Kripfganz]

Without a regression constant, the estimates will almost always be biased, potentially severely. The nocons option should only be used when there is a theoretical justification (e.g., all variables have been standardized, or all variables are in first differences and do not follow a time trend). It is not an option to experiment with. The nocons option will certainly not help with heteroskedasticity or serial correlation. If the problem "disappears", that is just a reflection of the uselessness of the results without a constant.

In short, do not use the nocons option.

[ahmed mohamed ak]

What are the conditions and steps for application Generalized method of moments estimation of linear dynamic panel data models?

[Sebastian Kripfganz]

What are the conditions and steps for application Generalized method of moments estimation of linear dynamic panel data models?

This is posted in the wrong topic. The following might help:

• https://www.statalist.org/forums/for...ent-conditions
• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.

[ahmed mohamed ak]

What are the conditions and steps for applying the CS - ARDL model?

[Sebastian Kripfganz]

This topic is about the ardl command for time series analysis. For CS-ARDL estimation with panel data, have a look at the xtdcce2 command. You can find various discussions about it here on Statalist and supporting material elsewhere on the web.

[ahmed mohamed ak]

If the slopes are heterogeneous and there is cross-sectional dependence, what is the appropriate model?

[Sebastian Kripfganz]

See my previous answer:

This topic is about the ardl command for time series analysis. For CS-ARDL estimation with panel data, have a look at the xtdcce2 command. You can find various discussions about it here on Statalist and supporting material elsewhere on the web.

[ahmed mohamed ak]

If the sample size is less than 30 observations, some researchers apply the Bootstrap ARDL approach to improve inference reliability. What is your opinion on this practice?

[Sebastian Kripfganz]

It is not clear to me why the Bootstrap should yield reliable inference in such small samples. The Bootstrap is an asymptotically justified resampling procedure; if there is hardly any data to sample from, then the Bootstrap cannot do any magic. Confidence intervals might be narrower than those obtained from conventional standard errors, but that does not mean that they are more reliable.

The Bootstrap can be useful if there is reason to believe that the error term has a non-standard distribution, but it would still require a reasonably large data set to construct a bootstrap sample that accurately reflects this non-standard distribution.

[ahmed mohamed ak]

Should we apply the ARDL bounds testing approach proposed by Pesaran, Shin, and Smith (PSS) when the sample size is less than 30, without resorting to the Bootstrapped ARDL test for cointegration?