Esteemed Colleagues,
I have this knotty question about interpreting the results from a dynamic panel threshold in Stata and will appreciate helpful feedback from those who have used the XTHENREG command developed by Seo & Kim (2016) and Seo et al (2019) in Stata.
Following Seo & Kim (2016) and Seo et al (2019), I estimated this model:
xthenreg lngdp lngfcfgdp lnlabr lntrade lnmob
where lngfcfgdp is the transition and threshold variable.
I have attached the result obtained.
Threshold value is 2.79 = r
My challenge is this: which of these coefficients coded _d or _b correspond to higher or lower values at the threshold point (r).
My interpretation (not sure about correctness):
For instance, will I be correct to say: "a percentage change in mobile subscription causes a decline in economic growth by 0.449 when GFCFGDP value is above the threshold of 2.79".
Kindly guide me, please.
Thanks.
Ngozi
See these links:
https://www.statalist.org/forums/forum/general-stata-discussion/general/1495808-how-to-deal-with-xthenreg-for-dynamic-threshold-models
https://arxiv.org/pdf/1902.10318.pdf
Article Estimation of dynamic panel threshold model using Stata
Thanks Remal Al-Gounmeein for your response and contribution. Deeply appreciated.
I have read this article countless times in addition to Seo and Kim (2016) but the authors left out a HUGE knowledge gap when they wrote the "xthrenreg" command. They never explained what _b and _d stand for.
Also, in my search for answers, I have gone through the same link you shared. Everyone who commented on that post are unsure what _b and _d connote. Just as others mentioned in the post, I sent an email to the authors early this year for clarification but they never responded till date.
Thanks anyways but I am nowhere close to the answer I anticipated.
Kind regards,
Ngozi
Dear Ngozi,
Stata codes for analysis could be taxing really. I suggest you download more and recent journals on the issue, you will surely get a way forward.
All the best
Thanks a lot for the suggestion, Kehinde Mary Bello
...deeply appreciated!
The authors really left a huge gap in the literature with their nature of scientific reporting on the subject, in their papers. However, several papers that adopted the method have always assumed that the lower regime is the "b" and the upper regime is the "d". First, the assumption is based on the view that "b" comes before "d". Another insight is the fact that when their threshold models are estimated without accounting for these regimes, only "b" is shown.
I hope this is helpful, madam.
Thanks Dr. Segun Thompson Bolarinwa for the helpful contribution.
...appreciated!
@Ngozi ADELEYE ..Dr were you able to find a solution to your question ?
Hamid Muili,
I had to re-read Seo et al (2019) and also agree with Dr. Segun Thompson Bolarinwa's contribution which suggest that:
_b = lower regime of the threshold
_d = upper regime of the threshold
....so, this gives the answers to the query.
I should look at the greater than and less than signs to do the interpretation.
Ngozi, your interpretation is correct. I’m sure you’ve also taken note of Segun’s helpful clue.
You can read Ndoricimpa’s paper on “The Threshold Effects of Inflation on Economic Growth:…(2017)” for more helpful insights on how to report and interpret results from dynamic panel threshold models.
All the best.
Dear Dr. Ngozi ADELEYE
I expect you have already found a way to interpret the results. However, I suggest reading the original article "Threshold effects in non-dynamic panels: Estimation, testing, and inference" by Bruce E.Hansen. I hope this will be useful for many other researchers.
Please find the paper using this link: Article Threshold Effects in Non-Dynamic Panels: Estimation, Testing...
All the best !!!