Basics of Panel ARDL

There are several reasons for conducting a panel data analysis. Such as: (1) the main interest is the “group” and not the individual units in the group, which means that very little information is lost by taking the panel perspective; (2) the use of panel rather than time series data not only increases the total number of observations and their variation but also reduces the noise coming from the individual time series (heteroscedasticity not an issue in panel data analysis); (3) best suited where data availability is an issue particularly for developing countries where short time spans for variables are rampant, often insufficient for fitting time series regressions; (4) there is heterogeneity (differences) among units in the panel; (5) panel estimation techniques take these heterogeneity into account by allowing for subject-specific variables; and (6) suited for studying dynamic changes due to repeated cross-sectional observations. I have listed 10 steps to panel ARDL estimations. They are: (1) specify the model; (2) describe the data; (3) perform correlation analysis; (4) perform unit root tests; (5) determine the optimal lags for the model; (6) perform the Hausman test; (7) perform cointegration test (optional); (8) estimate the model (s); (9) perform causality tests (optional) and (10) perform diagnostics (optional).
Follow up with soft-notes and updates from CrunchEconometrix:
Website: cruncheconomet...
Blog: cruncheconomet...
Forum: cruncheconometr...
Facebook: / cruncheconometrix
KZbin Custom URL: / cruncheconometrix
Stata Videos Playlist: • (Stata13):Estimate and...
EViews Videos Playlist: • (EViews10):Interpret V...

[@CrunchEconometrix]

KZbin recently changed the way my content will be monetised. My channel now needs 1,000 subscribers. So it would be amazing if you show your support by both watching my videos and subscribing to my channel if you haven’t done so already. Monetising my videos allows me to invest back into the channel with some new equipment so this small gesture from you will be extremely huge for me. Many thanks for your support….CrunchEconometrix loves to teach, support my Channel with your subscription and sharing my videos with your cohorts.

[@bettytuhaise2686]

Thanks for that supportive heart and for being generous with your knowledge base. I have been greatly helped.

[@CrunchEconometrix]

Great to hear, Betty! 🥰

[@elinakim6349]

You are a blessed and gifted educator, Dr.Ngozi! you've been such a blessing. God bless you more!

[@CrunchEconometrix]

Hi Elina, I'm encouraged and humbled by your feedback. May God bless you. Please may I know from where (location) you are reaching me?

[@elinakim6349]

@@CrunchEconometrix amen! I'm from Kazakhstan 🇰🇿, but currently a foreign student in China 🇨🇳 doing my master's. I really mean what I've said previously. You've been like an answer for my prayers asking God to help and lead me through. With your clear instructions I've conquered successfully through bachelor's and graduated in 2018. Now this is my final mile in master's and you're again here leading me through it again. During bachelor's dealt with time series, now panel data😁 let me repeat myself, you're a blessing🙏

[@NihatGumus-gj9bm]

I just come up with your channel. As a teacher of Financial Econometrics, I strongly recommend your channel to my students. Thank you for your support to the "Sharing Economy".

[@CrunchEconometrix]

I appreciate your recommendation...thanks!

[@felixlagemann8109]

Great professor, thank you Dr. Ngozi!

[@CrunchEconometrix]

U're welcome, Felix...may I know from where (location) you are reaching me?

[@felixlagemann8109]

From Erasmus University in Rotterdam, Netherlands :)

[@CrunchEconometrix]

@@felixlagemann8109 Awesome! I will appreciate if you can spread the word about my videos to your students and academic community in Netherlands 🇳🇱! 💕 😊

[@rayeugene5090]

Prof your tutorials are always simplified, accurate and educative. Thank you

[@CrunchEconometrix]

Glad you like them...thanks!

[@abdullaigaucho653]

I just came across this video today and I believe your videos will have a tremendous benefit in my studies. Prof, I endorse your style of teaching.

[@CrunchEconometrix]

Glad it was helpful, Abdullai! ...and thanks for the encouraging feedback, deeply appreciated!

[@YonasKiflom-o2h]

Best lecture ever I have seen. Thank you Prof.

[@CrunchEconometrix]

Wow, thank you!

[@BrianJasonPonce]

Again, another job well done, Dr. Thank you for this highly informative and straightforward lesson!

[@CrunchEconometrix]

Thanks for the encouraging words and feedback, Brian!

[@katrinahtenjah9460]

You have been such a blessing in my econometrics journey. Thank you prof

[@CrunchEconometrix]

Thanks Katrinah, for the encouraging feedback. Deeply appreciated! ❤️🙏

[@mohammedalnour318]

Excellent performance, you may not know how is it helpful . Thanks Dr

[@CrunchEconometrix]

I always appreciate your encouraging feedback, Mohd. Deeply appreciated! 🙏

[@mohammedalnour318]

@@CrunchEconometrix Dr, I do appreciate your efforts. I have some questions, how can I investigate the impact of COVID-19 on global economy empirically? using the Panel ARDL Model?

[@CrunchEconometrix]

Mohd, COVID-19 time series data will be a great challenge given it is a current phenomenon. You can do a qualitative study, cross-sectional analysis or descriptive analysis if you have sufficient data.

[@mohammedalnour318]

@@CrunchEconometrix thanks Dr, do you think that the structural equation modelling (SEM) can be a better technique of analysis for such study?

[@CrunchEconometrix]

Can't say, honestly.

[@parfaitberi3974]

Thank you for this video.

[@CrunchEconometrix]

You are welcome, Sir.

[@edgarndani5238]

Love you. This is wat I was looking for. Thank you Ma

[@CrunchEconometrix]

U're welcome Edgar. I had no idea that this was the procedure you meant. Glad I could help in some little way...😊

[@rachelmacauley3818]

Excellent video. Thank you very much indeed.

[@CrunchEconometrix]

Thanks for the positive feedback, Rachel...I'll appreciate if you can help spread the word about my Channel by sharing my link and videos😍

[@SharafatAliPK]

You're doing a great job Adeleye

[@CrunchEconometrix]

Thanks for the positive feedback, appreciated 🙏

[@moonsafar5718]

Thank you very much for useful video.👍

[@CrunchEconometrix]

U're welcome, Moon Safar😊. Please may I know from where (location) you are reaching me?

[@moonsafar5718]

@@CrunchEconometrix I am from middle east.😍

[@moonsafar5718]

Could you direct me how can I estimate in panel data ( ARDL, OLS, J.J OR E.J).

[@CrunchEconometrix]

@@moonsafar5718 Watch the panel ARDL collection. I'll upload those of OLS in a few days. What's J. J and E. J?

[@hudamajid2842]

very informative video.

[@CrunchEconometrix]

Compliment is humbly taken, Huja! May I know from where (location) you are reaching me?

[@بخيتهجادكريم]

thankyou very much it enrich tutorial

[@CrunchEconometrix]

U're welcome, Sir 🥰

[@keeeabdnfjks6800]

Thank so muchh for helpíng.

[@CrunchEconometrix]

Happy to help! 🥰

[@hidayatullah7068]

Nice I like it's lactures.

[@CrunchEconometrix]

Thanks Hidayat, for the positive feedback. Please share my videos with your colleagues and may God bless you, amen!

[@hidayatullah7068]

@@CrunchEconometrix offcours

[@oumasstabdallah9355]

I miss you beautiful PROFESSOR. ❤❤❤❤❤

[@CrunchEconometrix]

Thanks so much, Sir! 🙏

[@busayovictorosuntuyi4716]

Thanks for you clear explanation Madam

[@CrunchEconometrix]

U're welcome, Sir! Positive feedback is deeply appreciated!

[@ngusbekele2920]

thanks for your information DR. but what if the number of observation is greater than the the time period ( what will be the modal specification). I am eagerly looking for ward for your answer. thanks for your time...

[@CrunchEconometrix]

If N > T, you can deploy FE, RE, GMM, and Error Component Models. Thanks.

[@dhakaramkadel3671]

Awesome presentation. would you please see this? While conducting the ARDL model with four variables in Eviews, all the variables are non-stationary at level, but at the first difference, i.e., I(1), two variables, including the dependent variable, are stationary without taking their log, but two other independent variables are with log only. Should I need to use a log for all the variables in this situation?

[@CrunchEconometrix]

Thanks, Dhaka for your positive feedback. Deeply appreciated ☺️. Functional forms of a model is very crucial. Can greatly influence the outcome of your results. Take all logs and see what you get compare it to when you have mixed forms.

[@dhakaramkadel3671]

@@CrunchEconometrix Thank you very much.

[@RohiNkwama]

thanks Professor for the great video. I have a Question: can you run a Panel ARDL regressions with a mixture of variables, some at levels and some at first differences?

[@CrunchEconometrix]

Yes but it is important that the dependent variable is I(1).

[@onwejoshuachukwuma4395]

Thanks Dr for the Video Please what's your take on this What Econometric estimation can be use for panel small N (country) and large T( Time) when unit root is at level ? Again ma when one have a missing data in a panel can one interpolate and use the interpolated data for estimation. Hope to get feedback from you .

[@CrunchEconometrix]

Hi Jo, if all the variables are I(0) perform pooled OLS. I worry less about interpolating data. I use unbalanced panel data as it is. Thanks.

[@onwejoshuachukwuma4395]

@@CrunchEconometrix Thanks very helpful

[@enongenebetrand1119]

Much thanks our academic engine for the insight.

[@CrunchEconometrix]

Thanks for the encouraging words, Sir!🥰

[@moonsafar5718]

Thank you very much, you are super prof, could you prepare video talking about Panel ARDL but on Eviews ( on program Eviews), I do not know if that is possible.👍👍👍

[@CrunchEconometrix]

Hi Moon Safar, I'll do my best. Though, unlike Stata, the EViews interface is not very friendly for panel data analysis...and thanks for the compliments, humbly accepted😊❤️

[@aminaahmedalibelal5676]

One of your great videos. If my variables are cointegrated with Pedroni test, but the results of long run equations show insignificant variables. Does this mean cointegration is not present? Thanx

[@CrunchEconometrix]

Hi Amina, you can use the Pedroni results to validate cointegration.

[@josephtokpah8238]

Thanks very much for such an excellent explanation. Prof. I want to know whether there is a way to do the stability test using PMG ARDL?

[@CrunchEconometrix]

Thanks, Joseph for the encouraging feedback. Not sure about stability diagnostics in PMG. You may want to check out other online resources.

[@kejisunday296]

Well done Dr. Ngozi. Please, which of the following is best estimating spillover effects in a panel data anylysis? Is it System GMM or Panel ARDL.

[@CrunchEconometrix]

Hi Keji, none of these.

[@takrimas]

It is a wonderful lecture video. However, would you mind telling me how FMOLS and DOLS can be implemented on the ARDL framework?

[@CrunchEconometrix]

Hi Takrima, I have noted the suggestion but in the interim, you may need to check other online resources. Thanks.

[@themhamadi]

You are amazing teacher. Great professor, thank you Dr. Ngozi! When are you going to produce Vector error correction on Panel data with STATA?

[@CrunchEconometrix]

Thanks Moundhir, for the encouraging feedback. Deeply appreciated! At the moment, I have videos on PVAR-GMM published on my Teachable platform cruncheconometrix.teachable.com

[@dusandrljaca1492]

Dear, How to perform JB test and test for kurtosis and skewness in stata for panel data analysis? Is there any specific test for panel data or it is enough to set panel and specify model and then use classic tests (sktest) ?

[@CrunchEconometrix]

Dusan, I have not done the JB test using Stata but you can try out your suggestion.

[@kazikssss]

@CrunchEconometrix why didnt you mention that it is vital to check for cross-sectional dependence before moving to the unit root test ?

[@CrunchEconometrix]

Jan, now you know may I understand what your issues are?

[@kazikssss]

@CrunchEconometrix I'm just asking is there some kind of rationale behind why you haven't mentioned this ? As far as I know in order to conduct panel ardl you need to always check for cross sectional dependence.

[@CrunchEconometrix]

Jan, there's no rationale.

[@kazikssss]

@@CrunchEconometrix so if a cross sectional dependence is found the panel ardl is unreliable as far as i understand ?

[@CrunchEconometrix]

Jan, your query shows you are yet to fully understand the intricacies of pabel ARDL. I have always emphasised that video tutorials are insufficient due to situations like this. Please find the time to read published articles that used panel ARDL to support whatever you gained from watching my video. I have listed some papers at the end of the clips. Best regards.

[@md.mahfujurrahman864]

thanks...basically first 2 min awesome

[@CrunchEconometrix]

U're welcome! 💕

[@clarajoanjoachim2179]

Hello prof, I can't seem to understand the difference between subscript i and j in the generalized model. In the model, you showed that coefficient of lagged dependent variable has only subscript i, but in the explanation, there is both i and j. Could you explain that, please? thank you.

[@CrunchEconometrix]

Clara, I suggest you pick up any article that used panel ARDL and read up on the Methodology Section. That will help you understand the generalised equation.

[@jokosusilo9009]

Great explanation, ma'am professor, but I want to ask something That's the meaning of the ARDL model below of the notation for the sigma p and q? There is j=1 and j=0? Does j represent a unit or group sample from a cross-section or something else?

[@CrunchEconometrix]

Hi Joko, i have responded to this same query on another thread you posted it. They represent time specifications.

[@silarbiyoucef5888]

Excellent explanation, please how do we explain the significance of the parameters in cs-ardl model ?

[@CrunchEconometrix]

Hi Silarbi, results interpretations are mostly the same across different estimation techniques. So, give the usual interpretation.

[@emmanuelsenior1191]

Good display. Please how do I resolve the issue of near singular matrix aside dropping variables. I tried running panel ARDL estimation but I keep getting near singular matrix even after dropping almost all my independent variables

[@CrunchEconometrix]

Emmanuel, with panel ARDL the correlation has to do with the underlying lags of the independent variables. I suggest you use 0 lag for the regressors and 1 lag for the lagged depvar just as I did in the video. The lag structure is (1 0 0 0).

[@emmanuelsenior1191]

@@CrunchEconometrix thank you very much Dr. for the response I would apply that

[@emmanuelsenior1191]

I tried that but there's no option for 0 lag for regressors so I set both for 1. Still near singular matrix

[@CrunchEconometrix]

Emmanuel, are you using EViews or Stata?

[@emmanuelsenior1191]

@@CrunchEconometrix Dr. please am using E-views

[@reinagiovanni]

Hi Prof! What if i have a variable that requires the second difference?

[@CrunchEconometrix]

With I(2) variable, not sure if panel ARDL is a suitable technique.

[@bbouchra1000]

Thank you so much. It was so helpful

[@CrunchEconometrix]

U're very welcome. Please may I know from where (location) you are reaching me?

[@bbouchra1000]

@@CrunchEconometrix from Morocco

[@bbouchra1000]

@@CrunchEconometrix please one more question. Could we introduce control variables in this model in order to attenuate potential omitted variable bias ? If yes how could we do? Thank you in advance 🙏

[@CrunchEconometrix]

@بشرى **** Control variables are explanatory variables which can be included. Kindly watch the practical application to see the process. Thanks.

[@bbouchra1000]

@@CrunchEconometrix thank you so much .. we are learning a lot from your channel

[@dejeneadugna7534]

When our variables have different lag length ,how we specify the command while using ARDl_PMG? thank you

[@CrunchEconometrix]

Hi Degene, due to multicollinearity, I restrict to using 1,0,0,0 lag length.

[@moviesanddramakorea]

Prof, I ran a slope homegenity test for my variables after running a unit root test, i.e, I ran the slope homogeneity test on the I(1) of my variables since they were stationary at first difference, therefore I am going with PMG estimator, is this okay Prof?

[@CrunchEconometrix]

Hi Kezia, please read more about the slope homogeneity test to be certain that it's a prerequisite for the PMG analysis. Thanks

[@sabashah1160]

Thanku mam Please suggest some study material for panel ARDL

[@CrunchEconometrix]

Hi Saba, there are suggested references at the end of my panel ARDL videos. Thanks.

[@sabashah1160]

Ok mam Thanku for replying

[@elinakim6349]

Dr. Ngozi, please, would panel ARDL be applicable for a panel data set with existence of cross sectional dependence?

[@CrunchEconometrix]

Yes, because the econometric construct is in 1st difference.

[@elinakim6349]

@@CrunchEconometrix Thank you!

[@asfiabinteosman5303]

Professor, is it possible to run a panel data using 11 years data of 4 banks? Need your help. Please do answer

[@CrunchEconometrix]

Asifa, at least 30 years.

[@nausheensodhi803]

Thank you for this video. It is so clear and on point. However, I have to clarify if panel ARDL can be used for a dataset containing T=15 and N=19. Any input on this will be highly appreciated.

[@CrunchEconometrix]

No Nausheen...use FE, RE, POLS, and GMM.

[@lakurelabpradeeppanthi2257]

Is it necessary to have ECT always negative and significant? What if we get negative ECT coefficient but insignificant? IF ECT is not significant, but individual long-run coefficient is significant than can we consider them as long-run causality? What do you mean by joint causality? If ECT comes negative and significant but all individual regressors become insignificant then at that condition how should we conclude?

[@CrunchEconometrix]

Hi Pradeep, these are very easy to interpret. An insignificant coefficient has no impact on the outcome variable. Long-run causality can be inferred when the long-run coeffs are significant. Joint causality can be inferred when the ECT is significant. Kindly watch my causality videos for more understanding.

[@donasp5391]

Thank you! Great resource! I subscribed! 1. If I have 368 companies and period is 2010-2018, I cannot use ARDL if I understood correctly. What would you suggest? I would appreciate if you can suggest me an appropriate model and maybe a book for more information about my case. 2. If I would have 2 groups of countries (highly innovative and slightly innovative), could I use ADLR?

[@CrunchEconometrix]

Hi Miha, thanks for your subscription. I am grateful. 1) Since you have a short panel where N > T, panel ARDL is not applicable. Use FE, RE, and GMM. Kindly watch my videos on those techniques. 2) Watch my video on "Tips to building a panel data". Kind regards.

[@donasp5391]

Thank you so much! ☺️ Since I have missing data, could I use ML too? I was thinking GMM+ML. Kind regards.

[@CrunchEconometrix]

The "orthogonal" option takes care of missing values.

[@donasp5391]

Thank you very much! I am grateful for your advice. Have a wonderful day!

[@ngusbekele2920]

is that possible to use panel ARDL if the number of observation(N) is greater than the time period (T)

[@CrunchEconometrix]

To the best of my knowledge, NO.

[@fulyaozorhan2900]

Do you have a video on panel ARDL estimation by using eviews? I cannot find any in your channel.(There is only one with the STATA version.) Thanks

[@CrunchEconometrix]

Hi Fulya, the EViews platform is a bit complicated for me regarding panel data analysis. I'm still learning the interface. Once I'm confident, I'll replicate all my Stata panel data videos. Thanks for watching and keep sharing too! 💕 😊

[@moisedjepangkouamo469]

great and helpfull Dr. nevertheless can we have a playlists to how run a GMM model on eviews? and Panel ARDL on eviews. thanks !!!

[@CrunchEconometrix]

Kindly check other online resources. Thanks.

[@yoursjeffery]

Good day ma, Please how do I deal with panel data where variables have different orders of integration as inferred by the unit root test-I(0) (stationary at level), I(1) (stationary after first differencing), and I(2) (stationary after second differencing)?

[@CrunchEconometrix]

You can deploy PCSE and FGLS techniques.

[@yoursjeffery]

@@CrunchEconometrix thank you ma

[@MrWhy23]

should i use original data or differenced data in testing for ardl model or ardl bound testing as my all variables become stationary at I(1)?

[@CrunchEconometrix]

Watch my videos to see what I did. thanks.

[@lingchen5238]

Dr. Ngozi, thank you for your video,it's really very helpful,but i want to know ,do you have the NARDL model teaching video,looking forward to your reply,thank u

[@CrunchEconometrix]

Thanks for the positive feedback, Ling. But I don't have any videos on NARDL.

[@lingchen5238]

@@CrunchEconometrix OK ，Thank you

[@dusandrljaca1492]

Is it possible to use Kao test of cointegration in ARDL panel? Is it required with Pedroni test of cointegration that all statistics have to be lower than 0.05 in order to conclude that there is cointegration in panel data? Is it required to use some test for autocorrelation in ARDL panel model?

[@CrunchEconometrix]

Dusan, my practical panel ARDL videos address some of your questions. Support that with reading published articles that used the technique. Thanks.

[@oloyedeobagbuwa3790]

Dear Dr, I need your counsel again. I have panel data with large N (181 firms) and small T (16) and I want to confirm if I can use P ARDL. I will equally appreciate it if you can suggest possible models. I am looking at the effects of independent variables on the dependent variable both in the short and long run. My variables are stationary at level.

[@CrunchEconometrix]

Hi Oloyede, since N > T, applicable techniques are pooled OLS, LSDV, FE, RE, and GMM. Panel ARDL is applicable when N < T. Thanks.

[@nawalin12]

Thank you for sharing Dr.. I have a question, since panel ARDL is about long-run and short-run estimation, how to specify short-run and long-run model? thank you.

[@CrunchEconometrix]

Hi Nawalin, get any published article on panel ARDL to see how the model is specified.

[@nawalin12]

@@CrunchEconometrix Tq hank you for the quick respond, I did collect many of journals on panel ARDL, all of them just specify general panel ARDL equation and the reparameterized form, none of them specify clearly on long run and short run model.

[@CrunchEconometrix]

Then get the Pesaran, Shin and Smith (1995, 1997, 1999) papers...available on the internet.

[@nawalin12]

@@CrunchEconometrix I did, I have all of them, and make it as my main sources, read it many times.. but no explanation on short run and long run model.

[@CrunchEconometrix]

Read again. They developed the model. They are the ones we cite for panel ARDL. Adapt what they did to yours.

[@mehmetakyol4332]

Dear professor, I have a trouble about cointegration test. In my panel data variables depvar is I(1) and indep var is I(0). Can I use Westerlund cointegration test before performing pmg and mg? In some articles it implies that in that kind of stiation durbin hausman cointegration or bound test (f test) have to use to determine cointegraiton. I am really confused. Could you please kindly response . Thank you very much

[@CrunchEconometrix]

Hi Mehmet, you may follow the explanations I gave regarding cointegration in panel ARDL. Thanks.

[@HibaWorld]

Thank you for this informative video. I have a question. I have T>N (T=168 and N=28). My data is an unbalanced panel with one independent variable non-stationary and stationary after first differencing. I used PMG which turns out to be good after the Hausman test. However, I read that PMG doesn't account for cross-sectional dependencies and the option should be xtdcce2 model as it can test for cross-sectional dependencies using xtcd2. It should work with unbalanced data but it's not working. Wondering do you have any suggestion?

[@CrunchEconometrix]

Hi Nadia, thanks for the encouraging feedback. Deeply appreciated! You are right with all points raised. Why is the xtdcce2 syntax not working?

[@HibaWorld]

​@@CrunchEconometrix Hi it worked I had a row in my data which had missing value so after dropping it everything working.

[@aminaahmedalibelal5676]

Hi. Can we use ARDL in case of LARGE panel data?

[@CrunchEconometrix]

Hi Amina, kindly watch my videos on Panel ARDL for guidance. Thanks

[@lakurelabpradeeppanthi2257]

why the cointegration test is optional in Panel ARDL?

[@CrunchEconometrix]

Because the significance of the ECT evidences cointegration.

[@emregokceli5087]

Hi Prof. Thank you for the video. I have just a quick question. Should the dependent variable be non-stationary at level and stationary at I (1) to be able to use Panle ARDL? Thank you.

[@CrunchEconometrix]

Hi Emre, I should what is required in these panel ARDL videos.

[@gynendrabhandari1106]

please provide the literature source for the ardl panel data with T>>N

[@CrunchEconometrix]

Hi Gyendra, I listed references at the end of the video.

[@gynendrabhandari1106]

@@CrunchEconometrix thanks

[@emregokceli5087]

Hi Sir, if N>T, then could l use panel ARDL ? Thanks

[@CrunchEconometrix]

Hi Emre, no.

[@emregokceli5087]

@@CrunchEconometrix thank your for your reply. Could you suggest a method for panel data in which N>T ?

[@CrunchEconometrix]

Hi Emre, FE, RE, GMM, IV-GMM, PSCC, and Pooled OLS.

[@moonsafar5718]

Hi prof, I have panel data T=340, N= 11, There is cross section dependency error and the stationary of variables is mix i(1) i(0), have 5 independant variabels and one dep varabels, can I run Panel ARDL or not, thanks in advanced ?

[@CrunchEconometrix]

You mean that you have 340 years?

[@moonsafar5718]

@@CrunchEconometrix thank you prof.10 years with four quarters.

[@CrunchEconometrix]

Some augmentation of the panel ARDL estimators account for CSD. You can check out the other online resources.

[@moonsafar5718]

@@CrunchEconometrix could you please direct me to specific research ❤❤❤

[@CrunchEconometrix]

There are several articles. Just do a Google search.

[@yassinyahia2453]

Dear Prof. Ngozi, thanks for your nice job. Do you have video on how to perform pooled mean group estimator in stata?

[@CrunchEconometrix]

Yes Yassin, that is what I did in this panel ARDL series. Kindly watch to get more information supported by reading references indicated at the end of the videos.

[@yassinyahia2453]

@@CrunchEconometrix Dear Prof. Ngozi. Thnx very much for your prompt response.

[@busaritajudeen9168]

You are amazing teacher. When are you going to produce Random Effect and Fixed Effect on Panel data with STATA? You be Amazing mother too

[@CrunchEconometrix]

Thanks for the kind words, Busari. I do have those videos. Check out the Stata Panel Data Playlists. Thanks 😊

[@evaggeliasiopi4746]

Dear Professor i have a panel data with with N> T, N=16,T=1 and i want to use 6 independent variables and 2 dummies wicth is the best model can i use ?Thanks in advance.

[@CrunchEconometrix]

Eva, you can decide after watching my videos on fixed and random effects, and GMM. But T = 1?

[@evaggeliasiopi4746]

@@CrunchEconometrix Τ=11 sorry thanks a lot

[@ghazalaafzal6049]

hello, can you plz tell me , can we use ARDL model on Unbalanced panel data?

[@CrunchEconometrix]

Sure...

[@ghazalaafzal6049]

@@CrunchEconometrix and plz also tell me that can we use ARDL model on less than 20 observations?

[@ghazalaafzal6049]

means 20 years data

[@CrunchEconometrix]

You need at least 30 observations.

[@CrunchEconometrix]

You need 30 years data, at the least.

[@tundeomotehinse514]

If I have 42 firms with 10 years of performance. Can I use this model

[@CrunchEconometrix]

Hi Tunde, no you cannot. You may want to check out my videos on FE, RE and GMM. Thanks

[@tundeomotehinse514]

What is the solution?

[@tundeomotehinse514]

But you used 13yrs and 143 countries, why is my own not appropriate?

[@CrunchEconometrix]

No, I didn't. This clip is about Panel ARDL.

[@CrunchEconometrix]

@@tundeomotehinse514 I responded to you on suggested techniques.

[@yuldoshboysobirov9580]

Hello, I have several questions. 1. Can panel ARDL PMG approach automatically defines the lags or should we employ lag selection? 2. Can we employ Panel ARDL without cointegration tests? Or are they mandatory? Thanks for your responses!

[@CrunchEconometrix]

Hi Yuldoshboy, thanks for the queries. I addressed them all in the hands-on videos. Kindly watch them, thanks.

[@vheektorig1444]

Hi prof... Your videos are highly educative. Thanks alot. I've subscribed. Please, I need ur urgent opinion... If I have my N to be 20firms and T to be 6years with 3 independent and 3 dependent variables, what's the best methodology under panel to use... Referral materials to assist me to build panel models in this line would be appreciated. Thanks in anticipation ma.

[@CrunchEconometrix]

Hi Vic, thanks for the positive feedback and your subscription. Deeply appreciated! Watch my videos on Hausman test, FE and RE, GMM, Error Component Models and Sub-Samples Analysis. Support them with references provided at the end of the videos. Thanks.

[@vheektorig1444]

@@CrunchEconometrix Thanks alot ma

[@akhliddinismailov3766]

Professor, how to include the lagged values to panel ARDL model?

[@CrunchEconometrix]

Watch the entire series. I showed what to do.

[@megadwicahyani3370]

Hello thank you, your video is very helpful. But I want to ask, so if we use heterogeneous panel data we can’t do the panel ardl? Because the data can’t be estimated in random effects while the panel ardl requires a hausman test?

[@CrunchEconometrix]

Hi Mega thanks for watching my video but I'm not quite getting your query because heterogeneous panel data is estimated by panel ARDL. Unless I'm missing something in your query.

[@ndouniamaonionguivanbrenta8618]

Thank you very much Sir. How to take the Hausman test in an ARDL? To my knowledge, the Hausman test is done with MCO. However, ARDL is another method that we use in spite of MCO. And I think the use of the Hausman test has no place in the ARDL model. And thank you for your answer Sir.

[@CrunchEconometrix]

Ndouniama, it is advisable to watch my videos before posting your queries which are now becoming incessantly irrelevant. I have videos on ARDL and also on the Hausman test. Watch them, read the listed references at the end of the videos and STOP 🛑 confusing yourself and anyone reading your queries. This is my honest advice to you. Take them to enable you get the best out of my KZbin Channel. Please know that I won't respond to any irrelevant query from you. Thanks.

[@ndouniamaonionguivanbrenta8618]

@@CrunchEconometrix Thanks Sir!

[@afreenessani9534]

Hi, I really liked your video of importing panel data. However, my concern is to learn how to import time series data in eviews, I.e. is it the same way or is it a bit different? Also, I need to run longitudinal causality on that data, can you help me a bit in that matter? Thanks in advance. Love, Afreen.

[@CrunchEconometrix]

Hi Afreen, thanks for giving my video a good pass😊...and yes I have a video on how to import time series data into EViews. Simply click on the EViews Playlist to search for the video. I also have videos on causality in Stata and EViews you can check them out. It's basically the same procedure and interpretation.

[@kubraylmaz5021]

hello can you help me? hausman test conclusion = chi2(10) = (b-B)'[(V_b-V_B)^(-1)](b-B) = -64.53 chi2 model fitted on these data fails to meet the asymptotic assumptions of the Hausman test; see suest for a generalized test what should I do? Good work

[@CrunchEconometrix]

Kübra, what code did you use?

[@kubraylmaz7174]

@@CrunchEconometrix hausman mg pmg, sigmamore

[@CrunchEconometrix]

Use this: hausman pmg mg, sigmamore.

[@kubraylmaz7174]

@@CrunchEconometrix thank you very useful good work

[@nssofod]

Thank you ma'am for such a nice and lucid explanation. I hv a query. I hv data where N=5, T=25 and 5 variables. I want find the relationship between growth and remittances. Is it worth to apply dynamic panel ardl model? My variables are stationary at 1st difference. Thanks in advance.

[@CrunchEconometrix]

Hi Debasis, I suggest that you increase the number of observations to over 30 given you have 5 variables to avoid loss of too many degrees of freedom.

[@nssofod]

@@CrunchEconometrix that means T>30?

[@CrunchEconometrix]

Yes.

[@nssofod]

@@CrunchEconometrix thanks a lot madam

[@rahmanakbi1905]

Hi Sir , I am Rahma, a sophomore at the University of Monastir, Mahdia, Tunisia. The problem is that my database includes 30 African countries from 1980 to 2017. But how can I estimate the ARDL dans Eviews 10 model?

[@CrunchEconometrix]

Hi Rahma, you have a large panel data. Unfortunately, I have panel ARDL videos in Stata and not EViews.

[@rahmanakbi1905]

@@CrunchEconometrix I watched his videos on STATA and EVIEWS. But I found that my database has missing variables.

[@CrunchEconometrix]

@@rahmanakbi1905 Aside from the missing values, you have a large panel data which makes panel ARDL (pARDL) technique to be not applicable. To use pARDL, T must be considerably larger than N.

[@mounaamari1752]

@@CrunchEconometrix You are excellent nd your videos are veryy very helpful please have you an idea about thresholds commands in stata 13?

[@CrunchEconometrix]

@@mounaamari1752 Thanks for the positive feedback and words of encouragement. Deeply appreciated! Not sure about the Threshold regression routine in Stata13 (never tried it yet). But you can find out by typing "help threshold" in the command window. Please may I know from where (location) you are reaching me?

[@asanteka.2403]

My Dear Prof, i have a doubt which just came to my mind. In an econometric specification can we interact a variable at levels with another one (in growth rate)?

[@CrunchEconometrix]

Hi Asanke, maybe. Could be possible if there is a relationship to investigate.

[@asanteka.2403]

@@CrunchEconometrix thanks my dear prof

[@berkpalandokenlier2358]

Hello there. First of all, thank you very much for this explanatory and useful video. You explained it very beautifully and clearly. There was a question I was wondering about myself. At least how many arguments are required for panel ARDL. Would it be 5 or 6 more? Or less? Thank you.

[@CrunchEconometrix]

Hi Berk, thanks for the positive feedback. What is implied by "arguments"?

[@berkpalandokenlier2358]

​@@CrunchEconometrix Sorry. I mean the number of explanatory (or independent variables. How many variables can be used at most? 6 or 7 variables possible?

[@CrunchEconometrix]

There is no direct response to this. You can find out when you run several simulations.

[@dr.elkhanrichardzada9855]

Do you know whether there is a stata command for panel NARDL?

[@CrunchEconometrix]

Hi Dr. Zada, perhaps...can't really say. I've never had cause to perform the procedure. May I know from where (location) you are reaching me?

[@chathuricaldera637]

Can I get fixed and random effect models to level data when data are stationary at first difference(Has unit root for level data)

[@CrunchEconometrix]

You cannot use RE or FE estimators on longitudinal panel data.

[@ndouniamaonionguivanbrenta8618]

Please Sir, why the diagnosis on the ARDL panel is not an obligation? I can't wait for your answer please Sir!

[@CrunchEconometrix]

I explained that in the respective clip.

[@ndouniamaonionguivanbrenta8618]

@@CrunchEconometrix Thank you for your positive feedback Sir. I am sure you are competent. Please can you send me the link of said video Sir? Sorry for the inconvenience Sir!

[@mohamedalhwary8841]

hello, How can I add a macro economic variables to set of firms data? (Panel data-Cross sectional).

[@CrunchEconometrix]

Hi Mohamed, add them the same way others are added. Watch my videos to see my data structure.

[@mohamedalhwary8841]

​@@CrunchEconometrix Thanks Dr for your reply, but Isn't there any problem causes by repeating the same macro inputs for the firms?

[@CrunchEconometrix]

You only want to add explanatory variables, not so?

[@mohamedalhwary8841]

​@@CrunchEconometrix No doctor, I want to add a new column, to the variable of each company, related to the macro economy like the exchange rate, the exchange rate is repeated within the same period for all companies. Company A: Sales(a) - Profits(a) - Exchange rate (in the periode). Company b: Sales(b) - Profits(b) - Exchange rate (in the periode).

[@mohamedalhwary8841]

Thus, I have recurring variables within each company’s Panel data. These data are macroeconomics and create correlation between variables and cause Matrix error inside Eviews.

[@muhammadkamrankhan5766]

How many obervations need to apply Panel ARDL model. Time series ARDL need 30 observation.

[@CrunchEconometrix]

Hi Muhd, it's the same as in TS ARDL ...at least 30years obs

[@nssofod]

Dear madam, I hv a panel with t=27 and n=5. Having dynamic heterogeneous cross sectional dependence can I apply CS-ardl? Bcz in some papers I found this model is applied for large panels. If this is true which model is suitable to tackle all these issues of dynamic heterogeneity and CSD given that I hv both I(0) and I(1) but no I(2)? Plz comment.

[@CrunchEconometrix]

Hi Debasis, the technique you suggested is ok. Same for CCE-MG, CCE, DCCE-MG, AMG, GLS, and PCSE. I advise you do further readings on these.

[@nssofod]

@@CrunchEconometrix thanks a lot

[@nssofod]

@@CrunchEconometrix if possible plz make a video on cs-ardl and related issues

[@CrunchEconometrix]

It's on my to-do-list. Thanks!

[@nssofod]

Dear madam, My hausman test favouring MG but the results of PMG are better. I am applying panel ARDL. Is there any flaw in my model specification?

[@tareklakhloufi3000]

Hello, I need your advice, regarding the choice of the suitable method in this panel study T = 5 N = 6 Number of variables: 5 thank you in advance

[@CrunchEconometrix]

Hi Tarek, no meaningful panel analysis can be done with this except pooled OLS which conveys very little information. Either you make N excessively larger than T and perform FE, RE and GMM techniques or T excessively larger than N and perform panel ARDL. I have videos on these techniques, kindly watch them. May I know from where (location) you are reaching me?

[@tareklakhloufi3000]

@@CrunchEconometrix i m form morocco, with OLS can i know the effect of each N

[@CrunchEconometrix]

@@tareklakhloufi3000 Nah, you can't. That'll work under panel ARDL approach.

[@tareklakhloufi3000]

@@CrunchEconometrix what is the most suitable panel approach to these properietes: N: 6 T: 5

[@CrunchEconometrix]

@@tareklakhloufi3000 Kindly refer to my 1st response.

[@hudamajid2842]

i would like to ask can i use this test for my master thesis? i want to test if there is a relationship between poverty and economic growth? which test should i use. thank you

[@CrunchEconometrix]

Hi Huda, that will depend on whether it's a time series or panel study. Kindly check my Playlists because I have models and estimators that addresses both. Thanks!

[@edgarndani5238]

how do you clean and winsorize a panel data?

[@CrunchEconometrix]

Edgar Ndani What's winsorize?

[@yassinyahia2453]

Hi Prof. Ngozi. Thank you very much for your continuous and valuable videos. I current working paper has a PMG result. My dependent variable is the manufacturing value-added (%GDP) (MVA), while the independent variable is intra-export (. xtpmg d.lmva3 d.linex d.lhli d.lint d.lhdi d.lfd, lr(l.lmva3 linex lhli lint lhdi lfd) ec(ec) replace pmg). The intra-export (EXT) has a positive and significant effect on the MVA. However, when I used the GMM (xtabond2), or xtpcse, or xtgls, the coefficient of the independent var (EXT) was either negative and significant or positive but insignificant, which did not support the baseline result (positive and significant)!!!. Would you help?

[@CrunchEconometrix]

Yassin, it is not unexpected that different techniques yield different results. It is up to you to decide which way to go.

[@dr.sureshmago9211]

Hello Ma'am. I am examining relationship between FDI, GDP and Export on the panel data of 12 countries for 39 years. When I apply ARDL model on the log transformed data of 3 variables, eviews shows an error of "Near Singular matrix" . Please guide me, how can I get out of this problem. PLEASE DO THE NEEDFUL. Regards

[@CrunchEconometrix]

...and what is the needful?

[@dr.sureshmago9211]

@@CrunchEconometrix Ma'am, I mean, I need your help. I am in dilemma , which technique should I use? and How to tackle this "Non-Singular Matrix " error which I face while applying ARDL model in eview ?

[@CrunchEconometrix]

Regress the log of the depvar on the LEVEL of the explvars because multicollinearity occurs from the logs of the explvars.

[@dr.sureshmago9211]

@@CrunchEconometrix Yes Ma'am, I did the same but the problem still prevails.

[@dr.sureshmago9211]

​@@CrunchEconometrix Ma'am I did the same, but problem still exists. But then, I tried by changing measure of my variables, means, taking FDI as % of GDP (instead of FDI at constant price), Exports as % of GDP (instead of Exports at constant price) and GDP growth rate (instead of GDP at constant price) , the problem of "Near Singular Matrix" is solved.

[@abdelhadibenghalem1332]

Dear dr. Bosede Ngozi Adeleye, I'm a beginner Ph.D. student from Algeria and i really need your help, I am working on a balanced panel data which include: N: 8 & T: 12. my question is: What is the suitable model that I should use? please accept my sincere salutations.

[@CrunchEconometrix]

Hi Abdelhadi, you are watching the exact video series that relates to the suitable model. Watch all and follow my procedures, do not skip any. Thanks.

[@abdelhadibenghalem1332]

Dear@@CrunchEconometrix I Am watching right now thank you so much.

[@CrunchEconometrix]

Abdelhadi, I forgot to add that T must not be less than 30 to avoid small sample bias.

[@abdelhadibenghalem1332]

@@CrunchEconometrix thank you again in my case of study the maximum of T is 12 so please what would you propose to me?

[@CrunchEconometrix]

@@abdelhadibenghalem1332 You can convert to quarterly data to increase the sample size to 48 observations per country. Seek further online resources on how to do this.

[@noora3195]

very useful video .. thank you but I have a question why the diagnostic tests are optional ? I run PMG which was chooses according to Housman test and the results were significant and alien with my expectation .but still want to support my model with some diagnostic test . can you please help me with this ? shall I do diagnostic test for each country data ? what if I found heteroscadasity or serial correlation or cross dependence ? does this imply that model is not right

[@CrunchEconometrix]

Hi Noor, diagnostics are optional in the sense that in panel data analysis heteroscedasticity and serial correlation are not issues but you may test for CSD if you wish. I don't know how that is done so you can seek more online resources on how to proceed. Thanks.

[@Dr_Shiny]

Hats off Dear ma'am Professor, Need your help... I have T=41, N=16, i.e. T>N. 1- Data have both Heteroscadescity and Autocorrelation 2- From 7 Independent Variables, 4 variables are stationary at the level and 3 stationary at 1st difference 3- there is cross-sectional dependence Ma'am, Would you recommend Panel ARDL model in this scenario ?

[@CrunchEconometrix]

Rain-Walker, it is obvious that you have not watched any of my panel data videos. Otherwise, you'd know that your data structure is a misfit for panel ARDL. I'll advise you watch ALL my videos on TIPS TO BUILDING PANEL DATA, FIXED AND RANDOM EFFECTS, and GMM series for better understanding. Thanks

[@soufianemahjoubi5758]

If N>T. Ardl panel data estimate for this case or no??

[@CrunchEconometrix]

Hi Mahjoubi, as I explained in the video the answer is NO.

[@soufianemahjoubi5758]

@@CrunchEconometrix hello, please tell me how to estimate mean group estimator (MG), pooled mean group (PMG) and dynamic fixed effects (DFE) in eviews

[@CrunchEconometrix]

I'm yet to understand how to engage these in EViews. Stata is more robust for panel data analysis.

[@soufianemahjoubi5758]

@@CrunchEconometrix you say on short ardl panel data we can use large N and Small T

[@CrunchEconometrix]

No, I didn't say that. Watch my panel ARDL videos again. I gave very clear explanations.

[@muhamamdmoeezraza2773]

need a guideline in estimation of panel data

[@CrunchEconometrix]

Hi Muhd, I have videos on panel data analysis using Stata and EViews. Kindly watch them for better understanding. Thanks.

[@rahmanakbi1905]

Hello sir; I am a student in 2 years of Master in Research of the Faculty of Economics and Management of Mahdia. I have an annual database of 30 countries from 1980 to 2017. I apply the ARDL approach for this database. My problem is that the command of the ARDL is not correct (the command I use: forval i = 1/30 { ardl my variables, maxlags (p, q ......) matrix list e (delay) di }) The results are "no observations" why What is the problem with this command? help me STATA 15

[@CrunchEconometrix]

Hi Rahma, for panel ARDL, N must be very small. So, reduce the 30 countries to about 10...and re-estimate the model. May I know from where (location) you are reaching me?