Beloved guest/subscriber, you have discovered my amazing TH-cam Channel tailored specifically for you and other beginners and intermediate users. Please do not keep me to yourself (lol). Kindly share my videos and links with your students, colleagues and academic community so that they too can SUBSCRIBE and learn with ease….and for the global community to be aware that applied econometrics can be simplified. My teaching approach is very practical. I adopt a do-as-I-do style. Many thanks to those who have supported me by telling others. Once again, CrunchEconometrix loves to teach, support my Channel with your subscription, likes, feedbacks and sharing my videos with your cohorts. Follow me on Facebook, Twitter and Reddit. Love you all, greatly!!!
Thank you so much for always bringing so much light and clarity, madam. God bless you! Very grateful for what you are doing. Following and learning from you from my bachelors. Currently doing masters by God's grace. Thank you! Your channel is a channel of blessing for me :)
Thanks Elina, for the positive and encouraging words. I will continue to give my best to the academic community. May I know from where (location) you are reaching me?
Hi Rokas, I am humbled and encouraged by your positive feedback. I hope to do more to help as many that are willing to learn..may I know from where (location) you are reaching me?
Anthony Atiirimbey thanks Anthony for the kind words and please let your social media community and academic community know about my TH-cam channel.. Kindly share😉
Hello Dr. Adeleye, how are you? Hope you're in good health. I thank you for your tutorial videos on simultaneous equations on time series. These help me a lot in understanding the methodologies that can be used in my research. Apart from these, I would like also to request for tutorial video about Non-linear ARDL (NARDL) model. Thanks a lot in advance.
Thnk you Prof for your thorough tutorils. Please I would like to request for a complete video on how to construct the recent Dynamic ARDL model in both stata nd eviews
Thanks for your great videos..could you please tell me the difference between CUSUM and CUSUM Squared? In my model, the CUSUM seems to go outside of the boundaries but stays well inside with CUSUM squared. I don't understand or know how to explain that.
Hi Mohd, I will always advise students to engage the literature for in-depth knowledge. My advice is that you search the literature for the differences between both and which of them is best is establishing model stability. Thanks.
Hi Chime, thanks for the encouraging feedback. Deeply appreciated! It is the coefficient labeled as "ec". Please may I know from where (location) you are reaching me?
This video is awesome!Im so thankfull for this Dr. Adeleye. But pardon me, I use "ardl y x" syntax on Stata 15, but it commqnd unrecognized, would you help me to explain it why please?Thank you very much🙏
Hi Syahnanda, thanks for the encouraging feedback. Deeply appreciated! It appears you are yet to install the ARDL syntax. Type "help ARDL" in the Command Window and follow the Stata prompts to install the syntax.
@@CrunchEconometrix Dr.Adeleye, pardon me, I have one more question please, In ardl model, in this video we looking for the maximum lag of each variable first then looking for the stationary of each variabel in their optimum lags. Why dont we looking for each variables stationary first then looking for their optimum lag Dr? Thank you very much Dr. Adeleye🙏
Hi Benson, thanks for kind words...deeply appreciated. Kindly know that due to abuse and unethical conduct, Stata dofiles used in my videos are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com/view/datashop.php The files don't cost much just a token to maintain my website. Thanks for your understanding and patronage.
as your case, we can conclude that the dummy variable is not statistically significant in the short run, right? If in my rerearch, the dummy variable is at lag 1, the result will be in dummy D1. So can (dummy) diff.results be accepted?
Dear Bosede Ngozi Adeleye I would be grateful if you could answer my inquiries regarding stationarity of the variables used in ARDL model.. usually when we have a non stationary variable, we do first differencing and if it becomes stationary we then regress the differenced variable.. why in ARDL we use the variables without differences even if it is not stationary? thanks alot
Dear Bosede Ngozi Adeleye First of all, take my cordial love. I hope you are well. After watching your all videos related to the ARDL model, I would like to ask you several questions regarding ARDL, Dummy Variables,, Structural Break, ADF Test, Lag Selection, and so on. 1. For selecting lag, we command varsoc [variable name] (for a lag of a single variable) and for regression varsoc [variable names] (for a lag of a regression model). If varsoc shows variable will take lag (1), then, we need to test Augmented Dickey-Fuller (ADF) for stationary checking with the same lag, I mean lag (1). Then the result should be stationary or non-stationary. If varsoc lag(1) and ADF lag(1) is non-stationary, then, what will we do? If varsoc lag(1) and ADF lag(2) is stationary, then, is it acceptable? If varsoc lag(1) and ADF lag(3) is stationary, then, is it acceptable? My concern is here, can we make stationary by taking more lags than the varsoc showed lags? Increasing lags more is better than conducting first difference or not? 2. After watching videos in youtube (your and others), for a structural break, you suggest ghansen test and another suggest zandrews test. As far as, I understand ghansen test deals with dummy variables, I mean adding dummy variables in the model and increase the stability of the model. My concern is here can we perform ADF for structural break data or do we need to use zandrews test? or just conducting ghansen command to add dummy variable and solving structural break issue? If we use ghansen command and add dummy variables, do we need to test ADF or zandrews? or just ghansen test is enough for stationary and dummy and unit root test and cointegration test? I need your cooperation. I am looking forward to getting in touch with you. Best Regards Md Rasel Miah
@@MdRaselMiah100 I can't. They are too many. Besides, I have videos that address almost all your queries. Please devote time to watch for better understanding. Thanks.
Thank you for the well-detailed explanation. I have an issue with determining lag length for my dependent variable. When I use varsoc for it, the criterions indicate no lag is optimal. Can you hypothesize why that is? It doesn't make sense to me that there would be no lag for a dependent variable in an ARDL model. Thank you.
Amazing teacher. Tank you very much, madam. I have a simple query as a beginner: is there any technique to choose the dependent variable, or is it based on theoretical justification, for example, C=f(y)+u?
Hello This model can be used if I have two variables of order i(1) In case of a structural change I can use G.H. Even though my variables are I(1) Thank you
Hello Professor, Thanks a lot for making such videos. It really helped a lot. However, I am facing some problem, I ran the ardl model and the ardl ec model. The ec model shows cointegration. The model passed all the diagnostic tests (Dwatson, B-Godfrey, Cusum). However, my main goal is to forecast the independent variable. I am not sure how to forecast with ardl model? Or i need to forecast with the ardl ec model since there is cointegration? Can you please make a video/or suggest me the codes to run forecast with ardl when there is cointegration? That would mean a lot. Thank you
Hi Mujtaba, thanks for the encouraging feedback. Deeply appreciated! But unfortunately I have no idea about ARDL forecasting. You may want to check other online resources.
Thank you. However, i am a bit confused about which eesults to report in my dissertation. My plan is to report the unconditional stata results followed by cointegration results and long run+ short run relationships. Do you think this approach is fine?
Another question is, which lags series should I list when I executed Bound Cointegration test. The diagnostic test also indicated there is a problem of serial correlation and Heteroscedescity. Please help to resolve all these issues. Great thanks in Advance!!!
Hello Sir... I am Shekhar Saroj, working on a paper based on time series data. While analysing, I am facing the problem of selecting appropriate tools and techniques. My humble request to you to help me out with my problems:- I am working with multiple linear regression models, where I have taken 1 dependent and 5 independent variables (Two major independent and 3 control variables). I wanted to check the main effect and interaction effect with this model. I checked the unit root test of all the variables, where I found that variables are stationary at different levels. 1. Dependent variable is stationary at the (0) level 2. One of my major independent is stationary at second level I(2) 3. All other variables are stationary at the first level. Question. 1. Which test would be appropriate to check the validation of the model? 2. Which test would be appropriate to check the long-run and short run association ? (Cointigratiion test)?? 3. Which test would be applied to check the casualty? 4. Are there any specific tools or techniques to check the interaction effect (moderation effect)? Please suggest any other relevant information which you find suitable for my study. Thanks & Regards
hello! i have a question. what should i do when the variables are still non stationary even after using the first difference in adf test? i hope you can help me. thanks! :)
Adding to the previous query, is there any technique in STATA, like in Eviews, to identify the structural break point? If so, how can I do it? And, should there be any such break, can I use the usual ADF test for unit root?
Hi Professor, I greatly enjoyed watching your video. I went through all steps and tests, and only found problems when running final diagnostic tests: bgodfrey-fail to reject the null of no serial correlation white's IM test-fail to reject the null of homoskedasticity jb test-reject the null of normality in distribution of residuals I try the cusum test, but get the error "Yvar has values outside the range [0,1]" . I am assuming my model is not stable due to the residuals variance.I have significant results otherwise, and already tested for stationarity as well. Do you have any advice? Not sure where to go from here. Thank you so much for your help!
Hi Heather, thanks for the encouraging feedback. Deeply appreciated! Since your model failed all diagnostics, you may test for structural break and start over. One measure does not fix an econometric problem so, watch my structural break videos on Chow test and Gregory-Hansen. Please may I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you, I will try testing for a structural break and watch your videos on Chow and Gregory-Hansen testing. I am an economist working out of California, USA. Thanks again!
Hi Once we have checked for stationarity and finds the differenced variables are stationary, should we use the differenced variable series or the original series?
Thanks. But in your example the variable lngini2 is non stationary at level and stationary after differencing. But in ardl model you used the non stationary variable. Is that so? Why we not used the differenced variable? Thanks again.
@@anilraj9908 That is because the ARDL algorithm will estimate the model using the 1st difference but the 1st difference of the series should be used if using the OLS algorithm.
My concern is here can we perform ADF for structural break data or do we need to use zandrews test? or just conducting ghansen command to add dummy variable and solving structural break issue? If we use ghansen command and add dummy variables, do we need to test ADF or zandrews? or just ghansen test is enough for stationary and dummy and unit root test and cointegration test?
@@CrunchEconometrix Dear Madam, Thank you very much for the video. It was very helpful. Another question that is related to this, is: Given that we have a structural break, can we use the usual ADF and PP to test for Unit Root? Thank you for your time.
Dear Ma'am i watched so many videos of you in which you have shown how to find long run cointegration exists but when it come to conclusion I do not find the video of how to calculate and interpret ARDL in case of short-term cointegration and how to calculate and interpret long-term bound test. and I do not see the robustness test and how to add multi dummy variable in case of multi structural breaks in the data and what is the ECM and how to interpret it?
Thanks so much Dr. Ngozi. With the Cusum square deviating from the 5% bound and then reverting back to stability, what comment should be made in my thesis with this result. Thank you?
would you please like to share the ARDL (bounds testing ) codes for short run and long run form in presence of a structural break of the variables (if you have the codes at your hand for any data set). i plan to create a dummy variable assuming value 1 for the years onward at which structural break in the dependent variable occurs . But there are also many breaks in the other variables used in the model. is it correct?
Thank you for the video, l would like to ask a question. I am doing research using ARDL bound test, l noticed that all of my variables are insignificant ie the p values of my variables coefficient is above 0.005. My question is, is it important to look for variables insignificant if you are looking at the long run or short run between variables. l also noticed that most researches only explain the part of F>L(1), they don't look at the p-value of the variables and in your videos, you didn't explain the significance of the variables, hence l am therefore asking if the variables p-value which tests for variable significance is important when interpreting ARDL bound test results especially when we are looking for cointegration between the variables. Thank you
Hi Ephiphania, the golden rule is that at least 50% of the coefficients must be statistically significant. Otherwise, noone will take your results seriously. So, in estimating an ARDL-ECM procedure, the significance of the ECT alone is not sufficient. Change some regressors and re-estimate the model. May I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you ma'am, may l know the procedures of reestimating my variables to make them significant. I am currently based in South Korea
Hi Professor, How do I get the lag when I am doing the ADF test for stationarity? i.e. - in your example, for the command "dfuller lndc, lags(1)", how did you get the value of the lag as 1? From the optimal lag results by varsoc? Thanks.
If varsoc lag(1) and ADF lag(1) is non-stationary, then, what will we do? If varsoc lag(1) and ADF lag(2) is stationary, then, is it acceptable? If varsoc lag(1) and ADF lag(3) is stationary, then, is it acceptable? My concern is here, can we make stationary by taking more lags than the varsoc showed lags? Increasing lags more is better than conducting the first difference or not?
Hi Ma'am, Good Day!! thanks for great contribution. I have dataset with 1 dependent and 9 explanatory variables of which 3 are dummy variables. According to step 2- the optimal lag value of the each variables are (3 1 1 2 1 0 0 1 1 1), Step 3-indicate that the 5 variables are stationary at I(0) and 5 are stationary at I(1). The question is, what maximum lag should I identify in step 4???.
@@CrunchEconometrix The value of the maximum lags identified is 3, but, if I use the maximum lags-3, I get an error response in the stata. How do I fix it?
Hi Tarek, because the ECT is only relevant in the long-run. It captures the speed of adjustment to long-run equilibrium...may I know from where (location) you are reaching me?
Dear crunch econometrix i have a question, how to get the ECM in a panel model in stata. I have a panel of 32 countries and i don't know how to get the ECM with my panel
Hi Omalams, for a start watch these videos: "Tips to building panel data" and "Basics of Panel ARDL" they will offer some information on panel data structures and the appropriate estimation techniques to deploy. Thanks.
@@CrunchEconometrix i have already watched it and i followed all the steps thanks for this video it was informative but i can't remember seeing ECM on panel or maybe i should use VAR or VECM. Thanks
@@CrunchEconometrix the haussman test suggested me the PMG model that i used but when i'm looking at the PMG model, there is ECT and i don't think ECT is ECM or i missed something sorry for bothering you
@@omalams9350 Your queries bear no reflection that you have watched my panel data videos because (1) this video is about time series ARDL, (2) watch the suggested videos and those in the panel ARDL series. They are in your best interest. (3) Endeavour to jot some points while at it and (4) read some selected references at the end of the clips.
@@CrunchEconometrix Thank you very much for your prompt response. However, I would much be appreciated if you maintain email service, if possible. website access using google is very challenging where I live.
@@yassinyahia2453 With a growing subscriber base, not possible to maintain an email service. Initially I was, but stopped. Could not keep up with the regular requests for data and dofiles...moreso, I am not charging any fees. So, I reason the arrangement must be a WIN-WI for both parties. I will advise you seek alternative arrangement with any of your academic colleague to access my site on your behalf. I have sent you the dofile, but this is a ONE-OFF. I won't accede to any subsequent email request (given the reasons mentioned)...thanks for your understanding.
![[안방1열 풀캠4K] 베이비몬스터 'DRIP' (BABYMONSTER FullCam)│@SBS Inkigayo 241110](http://i.ytimg.com/vi/AQ4RIghEvUs/mqdefault.jpg)
Beloved guest/subscriber, you have discovered my amazing TH-cam Channel tailored specifically for you and other beginners and intermediate users. Please do not keep me to yourself (lol). Kindly share my videos and links with your students, colleagues and academic community so that they too can SUBSCRIBE and learn with ease….and for the global community to be aware that applied econometrics can be simplified. My teaching approach is very practical. I adopt a do-as-I-do style. Many thanks to those who have supported me by telling others. Once again, CrunchEconometrix loves to teach, support my Channel with your subscription, likes, feedbacks and sharing my videos with your cohorts. Follow me on Facebook, Twitter and Reddit. Love you all, greatly!!!
All my thanks, Very grateful for what you are doing.
Thanks for your encouraging words, Basma...deeply appreciated!💖
Thank you so much for always bringing so much light and clarity, madam. God bless you! Very grateful for what you are doing. Following and learning from you from my bachelors. Currently doing masters by God's grace. Thank you! Your channel is a channel of blessing for me :)
Thanks Elina, for the positive and encouraging words. I will continue to give my best to the academic community. May I know from where (location) you are reaching me?
High quality content you have here! Thank you a lot.
Hi Rokas, I am humbled and encouraged by your positive feedback. I hope to do more to help as many that are willing to learn..may I know from where (location) you are reaching me?
Yes, no problem. I am from Lithuania.
Ngozi your lecture is amazing.
Best teacher ever!!!
I sent you an email for the do file but keeps bouncing back
Anthony Atiirimbey thanks Anthony for the kind words and please let your social media community and academic community know about my TH-cam channel.. Kindly share😉
Anthony Atiirimbey Maybe you're sending to a wrong email. My address is on my website... Once I get your request, I'll oblige.
Done! Just sent it
Anthony Atiirimbey Request seen. Dofile will be sent later today. Please acknowledge receipt when you have received it.
I really like your video, learned a lot, thank you!
Thanks, Sheleme, for your encouraging feedback. It is deeply appreciated! ❤️
Hello Dr. Adeleye, how are you? Hope you're in good health. I thank you for your tutorial videos on simultaneous equations on time series. These help me a lot in understanding the methodologies that can be used in my research. Apart from these, I would like also to request for tutorial video about Non-linear ARDL (NARDL) model. Thanks a lot in advance.
Thanks, Rosario for the encouraging feedback. Deeply appreciated 🙏. Videos on NARDL have been penned down. Thanks
Thnk you Prof for your thorough tutorils. Please I would like to request for a complete video on how to construct the recent Dynamic ARDL model in both stata nd eviews
Thanks for your suggestion. Duly noted. Thanks
Thanks for your great videos..could you please tell me the difference between CUSUM and CUSUM Squared? In my model, the CUSUM seems to go outside of the boundaries but stays well inside with CUSUM squared. I don't understand or know how to explain that.
Hi Mohd, I will always advise students to engage the literature for in-depth knowledge. My advice is that you search the literature for the differences between both and which of them is best is establishing model stability. Thanks.
Hi. What do you do if differentiating doesn’t make the variable stationary?
Hi Rahul, you can de-trend the series.
Thank so much ma for the video...pleasure which of them is the ECT value...am a little confused
Hi Chime, thanks for the encouraging feedback. Deeply appreciated! It is the coefficient labeled as "ec". Please may I know from where (location) you are reaching me?
You are amazing and extremely helpful. Big THANK YOU :)
U're welcome, Farah!...may I know from where (location) you are reaching me?
@@CrunchEconometrix I am Lebanese but I am studying in Paris, France.
@@fifita93 Awesome! Please spread the word about my videos...thanks! 💕
Thank you very much
keep it Up
You are welcome, Sir 🥰🙏
This video is awesome!Im so thankfull for this Dr. Adeleye. But pardon me, I use "ardl y x" syntax on Stata 15, but it commqnd unrecognized, would you help me to explain it why please?Thank you very much🙏
Hi Syahnanda, thanks for the encouraging feedback. Deeply appreciated!
It appears you are yet to install the ARDL syntax. Type "help ARDL" in the Command Window and follow the Stata prompts to install the syntax.
@@CrunchEconometrix thank you very much Dr. Adeleye for your respons, verry helpfull🙏
@@CrunchEconometrix Dr.Adeleye, pardon me, I have one more question please, In ardl model, in this video we looking for the maximum lag of each variable first then looking for the stationary of each variabel in their optimum lags. Why dont we looking for each variables stationary first then looking for their optimum lag Dr? Thank you very much Dr. Adeleye🙏
Syahnanda, you can try your idea too. Sounds ok🥰
Thanks, very grateful for your lecture. P
lease I need the do file
Hi Benson, thanks for kind words...deeply appreciated.
Kindly know that due to abuse and unethical conduct, Stata dofiles used in my videos are no longer free but available on my website upon payment.
Here's the link cruncheconometrix.com/view/datashop.php
The files don't cost much just a token to maintain my website.
Thanks for your understanding and patronage.
as your case, we can conclude that the dummy variable is not statistically significant in the short run, right?
If in my rerearch, the dummy variable is at lag 1, the result will be in dummy D1. So can (dummy) diff.results be accepted?
Hi Nguyen, there is no point lagging a dummy variable.
Dear Bosede Ngozi Adeleye
I would be grateful if you could answer my inquiries regarding stationarity of the variables used in ARDL model.. usually when we have a non stationary variable, we do first differencing and if it becomes stationary we then regress the differenced variable.. why in ARDL we use the variables without differences even if it is not stationary? thanks alot
Hi Mohammed, ARDL is robust to accommodate nonstationary variables. I advise you to read the Pesaran et al papers indicated at the end of the video.
Dear Bosede Ngozi Adeleye
First of all, take my cordial love. I hope you are well. After watching your all videos related to the ARDL model, I would like to ask you several questions regarding ARDL, Dummy Variables,, Structural Break, ADF Test, Lag Selection, and so on.
1. For selecting lag, we command varsoc [variable name] (for a lag of a single variable) and for regression varsoc [variable names] (for a lag of a regression model). If varsoc shows variable will take lag (1), then, we need to test Augmented Dickey-Fuller (ADF) for stationary checking with the same lag, I mean lag (1). Then the result should be stationary or non-stationary. If varsoc lag(1) and ADF lag(1) is non-stationary, then, what will we do? If varsoc lag(1) and ADF lag(2) is stationary, then, is it acceptable? If varsoc lag(1) and ADF lag(3) is stationary, then, is it acceptable? My concern is here, can we make stationary by taking more lags than the varsoc showed lags? Increasing lags more is better than conducting first difference or not?
2. After watching videos in youtube (your and others), for a structural break, you suggest ghansen test and another suggest zandrews test. As far as, I understand ghansen test deals with dummy variables, I mean adding dummy variables in the model and increase the stability of the model. My concern is here can we perform ADF for structural break data or do we need to use zandrews test? or just conducting ghansen command to add dummy variable and solving structural break issue? If we use ghansen command and add dummy variables, do we need to test ADF or zandrews? or just ghansen test is enough for stationary and dummy and unit root test and cointegration test?
I need your cooperation. I am looking forward to getting in touch with you.
Best Regards
Md Rasel Miah
Hi Rasel, this is just TOO LONG. Hence, I didn't bother to read. I respond ONLY to short and direct queries. Sorry.
@@CrunchEconometrix Questions are critical if possible then read and cooperate with me.
@@MdRaselMiah100 I can't. They are too many. Besides, I have videos that address almost all your queries. Please devote time to watch for better understanding. Thanks.
Dear sir if there are no cointegration ,how to use the command to run short run model
Hi Pheng, I showed what to do if there is no cointegration. You may need to watch this clip again in addition with my other ARDL videos. Thanks.
Thank you for the well-detailed explanation. I have an issue with determining lag length for my dependent variable. When I use varsoc for it, the criterions indicate no lag is optimal. Can you hypothesize why that is? It doesn't make sense to me that there would be no lag for a dependent variable in an ARDL model. Thank you.
Hi Iain, I find that difficult to comprehend but if that is the case use 1 lag and estimate your model.
Amazing teacher. Tank you very much, madam. I have a simple query as a beginner: is there any technique to choose the dependent variable, or is it based on theoretical justification, for example, C=f(y)+u?
Hi Rupon, your dependent variable is the essence of your research. Kindly read more about this from any textbook.
Hello
This model can be used if I have two variables of order i(1)
In case of a structural change I can use G.H. Even though my variables are I(1)
Thank you
Hii mam, I am still wondering how numbers for dum_lmdc and dum_dese for time span 2008-2015 are exist? Aren't they only 0 and 1?
Hi Nurmi, those multiplied by 0 equates to 0 and those multiplied by 1 retain there values. Try that with your variables.
Hello Professor,
Thanks a lot for making such videos. It really helped a lot.
However, I am facing some problem, I ran the ardl model and the ardl ec model. The ec model shows cointegration. The model passed all the diagnostic tests (Dwatson, B-Godfrey, Cusum). However, my main goal is to forecast the independent variable. I am not sure how to forecast with ardl model? Or i need to forecast with the ardl ec model since there is cointegration? Can you please make a video/or suggest me the codes to run forecast with ardl when there is cointegration? That would mean a lot.
Thank you
Hi Mujtaba, thanks for the encouraging feedback. Deeply appreciated! But unfortunately I have no idea about ARDL forecasting. You may want to check other online resources.
Thank you. However, i am a bit confused about which eesults to report in my dissertation. My plan is to report the unconditional stata results followed by cointegration results and long run+ short run relationships. Do you think this approach is fine?
Dear Madam, can you please tell me how I can interpret all the dummy variables?
Hi Mamun, kindly do a Google search on "dummy variables interpretations" for several online resources.
Thanks, Madam.
Thank you so much for this video. I want to know how can heteroscedasticity be corrected in ARDL model?. Thanks again
Hi Shania, you can take the logs of the series.
Still, it doesn't work. Any other suggestions pls. Thanks.
@@shanikarathnayake1072 At best that's what I do but you can also do more search regarding correcting heteroscedasticity online.
Thank you so much.
Another question is, which lags series should I list when I executed Bound Cointegration test. The diagnostic test also indicated there is a problem of serial correlation and Heteroscedescity. Please help to resolve all these issues. Great thanks in Advance!!!
This video shows you what to do... You may check online for additional information. Thanks.
Hello Sir...
I am Shekhar Saroj, working on a paper based on time series data. While analysing, I am facing the problem of selecting appropriate tools and techniques.
My humble request to you to help me out with my problems:-
I am working with multiple linear regression models, where I have taken 1 dependent and 5 independent variables (Two major independent and 3 control variables). I wanted to check the main effect and interaction effect with this model.
I checked the unit root test of all the variables, where I found that variables are stationary at different levels.
1. Dependent variable is stationary at the (0) level
2. One of my major independent is stationary at second level I(2)
3. All other variables are stationary at the first level.
Question.
1. Which test would be appropriate to check the validation of the model?
2. Which test would be appropriate to check the long-run and short run association ? (Cointigratiion test)??
3. Which test would be applied to check the casualty?
4. Are there any specific tools or techniques to check the interaction effect (moderation effect)?
Please suggest any other relevant information which you find suitable for my study.
Thanks & Regards
Hi Shekhar, kindly watch my time series videos including those on:
1) stationary test
2) moderation modelling.
...thanks
hello! i have a question. what should i do when the variables are still non stationary even after using the first difference in adf test? i hope you can help me. thanks! :)
Hi Micheal, better to drop it and use a closer proxy
Hi! May I know why you opted for lag (1) in checking for stationarity? I saw some sources using the optimal lag by AIC for this step. Thanks!
Hi Dino, I use 1 lags to avoid losing too many degrees of freedom.
Adding to the previous query, is there any technique in STATA, like in Eviews, to identify the structural break point? If so, how can I do it? And, should there be any such break, can I use the usual ADF test for unit root?
Watch my video on the Chow test.
You are wonderful
Thanks, Alexis for the encouraging feedback. Deeply appreciated!
Hi Professor, I greatly enjoyed watching your video. I went through all steps and tests, and only found problems when running final diagnostic tests:
bgodfrey-fail to reject the null of no serial correlation
white's IM test-fail to reject the null of homoskedasticity
jb test-reject the null of normality in distribution of residuals
I try the cusum test, but get the error "Yvar has values outside the range [0,1]"
. I am assuming my model is not stable due to the residuals variance.I have significant results otherwise, and already tested for stationarity as well. Do you have any advice? Not sure where to go from here.
Thank you so much for your help!
Hi Heather, thanks for the encouraging feedback. Deeply appreciated! Since your model failed all diagnostics, you may test for structural break and start over. One measure does not fix an econometric problem so, watch my structural break videos on Chow test and Gregory-Hansen. Please may I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you, I will try testing for a structural break and watch your videos on Chow and Gregory-Hansen testing. I am an economist working out of California, USA. Thanks again!
Hi
Once we have checked for stationarity and finds the differenced variables are stationary, should we use the differenced variable series or the original series?
Please follow steps as shown. Also, watch my other ARDL videos.
Thanks. But in your example the variable lngini2 is non stationary at level and stationary after differencing. But in ardl model you used the non stationary variable. Is that so? Why we not used the differenced variable? Thanks again.
@@anilraj9908 That is because the ARDL algorithm will estimate the model using the 1st difference but the 1st difference of the series should be used if using the OLS algorithm.
My concern is here can we perform ADF for structural break data or do we need to use zandrews test? or just conducting ghansen command to add dummy variable and solving structural break issue? If we use ghansen command and add dummy variables, do we need to test ADF or zandrews? or just ghansen test is enough for stationary and dummy and unit root test and cointegration test?
ADF is not applicable if there's a structural break. There are several test for unit root that are relevant: Z-A and so on.
Good morning Madam, How do I estimate an ARDL with regime shift?
You can check other online resources for this, thanks.
@@CrunchEconometrix Dear Madam, Thank you very much for the video. It was very helpful. Another question that is related to this, is: Given that we have a structural break, can we use the usual ADF and PP to test for Unit Root?
Thank you for your time.
No, you can't.
Dear Ma'am i watched so many videos of you in which you have shown how to find long run cointegration exists but when it come to conclusion I do not find the video of how to calculate and interpret ARDL in case of short-term cointegration and how to calculate and interpret long-term bound test. and I do not see the robustness test and how to add multi dummy variable in case of multi structural breaks in the data and what is the ECM and how to interpret it?
Abbas, I have those videos in my Stata Time Series Playlist. Browse through and watch them. Thanks.
@@CrunchEconometrix thank you ma'am but we wish to see it in eviews
I have them in EViews too. Search for ARDL videos in my EViews Time Series Playlist.
@@CrunchEconometrix would you please share the link in comment Thank you
But I just told you to watch my ARDL videos in EViews.
Thanks so much Dr. Ngozi. With the Cusum square deviating from the 5% bound and then reverting back to stability, what comment should be made in my thesis with this result. Thank you?
Hi Dennis, simple explanation: model was initially unstable and thereafter reverted to stability".
@@CrunchEconometrix Thank you
@@dennisbaidoo5995 No worries...please tell your cohorts about my YT Channel. I'll greatly appreciate it. Thanks!
@@CrunchEconometrix l will surely tell them. Thanks.
would you please like to share the ARDL (bounds testing ) codes for short run and long run form in presence of a structural break of the variables (if you have the codes at your hand for any data set).
i plan to create a dummy variable assuming value 1 for the years onward at which structural break in the dependent variable occurs
. But there are also many breaks in the other variables used in the model.
is it correct?
Hi Maruf, you can use dofile on Gregory-Hansen test for structural break using the ARDL model. It's available on my website. Thanks.
Thank you for the video, l would like to ask a question. I am doing research using ARDL bound test, l noticed that all of my variables are insignificant ie the p values of my variables coefficient is above 0.005. My question is, is it important to look for variables insignificant if you are looking at the long run or short run between variables.
l also noticed that most researches only explain the part of F>L(1), they don't look at the p-value of the variables and in your videos, you didn't explain the significance of the variables, hence l am therefore asking if the variables p-value which tests for variable significance is important when interpreting ARDL bound test results especially when we are looking for cointegration between the variables. Thank you
Hi Ephiphania, the golden rule is that at least 50% of the coefficients must be statistically significant. Otherwise, noone will take your results seriously. So, in estimating an ARDL-ECM procedure, the significance of the ECT alone is not sufficient. Change some regressors and re-estimate the model. May I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you ma'am, may l know the procedures of reestimating my variables to make them significant. I am currently based in South Korea
@@missstar8532 Same process.
when we are applying the ardl command shouldn't we use the differenced variables?
Not with the ARDL algorithm but with OLS.
Hi Professor, How do I get the lag when I am doing the ADF test for stationarity? i.e. - in your example, for the command "dfuller lndc, lags(1)", how did you get the value of the lag as 1? From the optimal lag results by varsoc? Thanks.
Yes Philip, lags used is obtained from the varsoc command on the respective variable.
Thank you so much Professor. It worked. And thank you for the video. It helped me a lot!
Glad to hear, Philip!👌
@@CrunchEconometrix , Ma'am, what should I do? if the lags value is other then 0 or 1. ???
If varsoc lag(1) and ADF lag(1) is non-stationary, then, what will we do? If varsoc lag(1) and ADF lag(2) is stationary, then, is it acceptable? If varsoc lag(1) and ADF lag(3) is stationary, then, is it acceptable? My concern is here, can we make stationary by taking more lags than the varsoc showed lags? Increasing lags more is better than conducting the first difference or not?
You can estimate your model with or without the optimal lags. Watch my video on the ADF procedure to know more. Thanks.
Hi Ma'am, Good Day!! thanks for great contribution. I have dataset with 1 dependent and 9 explanatory variables of which 3 are dummy variables. According to step 2- the optimal lag value of the each variables are (3 1 1 2 1 0 0 1 1 1), Step 3-indicate that the 5 variables are stationary at I(0) and 5 are stationary at I(1). The question is, what maximum lag should I identify in step 4???.
Hi Muhd, thanks for the positive feedback. Appreciated! Reduce your explvars. They are just TOO many and use the identified lags.
@@CrunchEconometrix The value of the maximum lags identified is 3, but, if I use the maximum lags-3, I get an error response in the stata. How do I fix it?
What's the error message?
@@CrunchEconometrix the error message is "# of lag permutations (786432) exceeds setting of 'maxcombs' (100000)"
That's a clear indication that your regressors are too many.
sr,pls help me?? stata with ARDL modle,can i send My data//???
Hi Habtamu, kindly watch my videos, follow the guidelines and analyze your data. Thanks.
why in the short term the probability of ECT is not mentioned
Hi Tarek, because the ECT is only relevant in the long-run. It captures the speed of adjustment to long-run equilibrium...may I know from where (location) you are reaching me?
@@CrunchEconometrix From morocco
why the command cusum6 gives me directly the test cousm square and not cusm test
Dear crunch econometrix i have a question, how to get the ECM in a panel model in stata. I have a panel of 32 countries and i don't know how to get the ECM with my panel
Hi Omalams, for a start watch these videos: "Tips to building panel data" and "Basics of Panel ARDL" they will offer some information on panel data structures and the appropriate estimation techniques to deploy. Thanks.
@@CrunchEconometrix i have already watched it and i followed all the steps thanks for this video it was informative but i can't remember seeing ECM on panel or maybe i should use VAR or VECM. Thanks
@@CrunchEconometrix the haussman test suggested me the PMG model that i used but when i'm looking at the PMG model, there is ECT and i don't think ECT is ECM or i missed something sorry for bothering you
@@omalams9350 Your queries bear no reflection that you have watched my panel data videos because (1) this video is about time series ARDL, (2) watch the suggested videos and those in the panel ARDL series. They are in your best interest. (3) Endeavour to jot some points while at it and (4) read some selected references at the end of the clips.
@@omalams9350 This thread is about time series ARDL not panel. Post your query on the appropriate thread.
In the first row sedu has a missing value but in case of dum_sedu first row has zero, why is it so?
Hi Sushrut, because dum_sedu for the 1st row is: 0*sedu = 0
Thank you very much, Dr for this valuable video. Would you tell us the guideline for model diagnostics. I have sent you an email for the do file.
Hi Yassin, the dofile contains the procedures as shown in the video. Get access on my website.
@@CrunchEconometrix Thank you very much for your prompt response. However, I would much be appreciated if you maintain email service, if possible. website access using google is very challenging where I live.
@@yassinyahia2453 With a growing subscriber base, not possible to maintain an email service. Initially I was, but stopped. Could not keep up with the regular requests for data and dofiles...moreso, I am not charging any fees. So, I reason the arrangement must be a WIN-WI for both parties. I will advise you seek alternative arrangement with any of your academic colleague to access my site on your behalf. I have sent you the dofile, but this is a ONE-OFF. I won't accede to any subsequent email request (given the reasons mentioned)...thanks for your understanding.
@@CrunchEconometrix Thank you very much, Dr Ngozi. I have received your email, and apologies for very late response
No problem, Yassin...I understand. U're welcome.