15:38 Choosing Lag Length 21:44 Reading VAR result 27:05 Granger Causality 35:41 Impulse 36:17 Variance Decomposition 38:19 Advantages of VAR Model 40:00 Problems with VAR 51:18 Why Stationarity Matters 1:09:08 Unit Root and Cointegration Tests 1:11:12 Engel-Granger Test 1:15:57 Granger Representation 1:20:56 VECM Johansen
This is the best econometrics video I have ever seen in my life !! Thank you for the best teaching method !! You should keep posting things more often !! You are doing a very very good job !!
Thank you for the class. You helped me a lot with this amazing explanation of a very difficult topic. I have read many books about the topic and now everything makes sense. I need to watch the video again, but I am pretty sure that, when finished, I will be ready to finish my last thesis chapter. Thank you again. Thank you a lot.
This is actually my first comment on UA-cam. I am currently writing my master's thesis in economics and have to learn VAR autodidactically - this is by far the best video I have found on UA-cam in this context. Hands-on and simply brilliant! Thank you very much & keep on teaching like this! Greetings from Germany!
Thank you very much for this awesome lecture. Just a few questions: 1. 18:25 : What is "root of the matrix"? Eigenvalues? Also, what condition would I need to fulfil if I had a, say, VAR(3)? 2. 01:06:00 When you discussed ECM and argued about usability of the OLS, it was probably wise to mention, that the integrating factor is also I(0). 3. What about VECM models with more than 2 variables and higher dimensions (like, 3+ lags)? I guess we can get more than one integrating factor then? What about condition on the pi-matrix? Will there be just 1 pi-matrix or more than 1?
For your question 1: I think the root of the matrix he mentioned are alpha and beta values of Yt-1 matrix, which is the coefficient when you fit the VAR(1) model. Similarly, when fitting VAR(3), there will be 2 additional A matrces (let's say A2 for lag 2 and A3 for lag 3). These matrices also need to be stable. The reason behind this is to ensure the variables in the model are stationary, which is consistent with the VAR model assumption (not too sure) For question 2: I dont think using OLS necessarily mean the integrating factor is I(0). I think when he means is by differencing the 2 series, we now have 2 stationary series, whcih mean variance do not change over time and thus can use OLS. VECM with more than 3 variables, lets say 3, the pi-matrix will be 3 by 3 matrix. VECM with lags more than 1, let's say 2, there will be 2 pi-matrix. Correct me if im wrong
thank you, sir... you have a nice way of explaining each and everything in a topic .this is the very first lecture on youtube from which I learn the VAR method, you explained this topic in a very easy way and this lecture will help me in my research work. thanks again. M.phil Student university of Peshawar, Pakistan
Hanomics I have one question Sir I have seven variables VAR Model, all endogenous. Should I do lag selection before stationerity or after stationerity at first difference...............my data is monthly data with 72 observations and I tried lag selection at first diff and AIC tells me 7 lags. The model does not become stable when I proceed for fcast and Irf analysis........kindly help me sir whether I shiuld select lags before I do stationerity at first diff. Thank you in advance!
Thank you so much for this wonderful lecture. I was really helpful in understanding VAR for the purpose of my thesis. I am currently doing a panel VAR analysis for 6 countries and have 6 variables. The problem I am experiencing is that, although all my variables are stationary at I(0) (using ADF, PP, LLC and IPS unit root tests), when I do the AR Roots test graph and table (using eviews 10) I find that my VAR is not stable regardless of the number of lags I use. What could be the problem and how can I fix it? I would like to interpret my Impulse Response Functions and Variance Decomposition. I would greatly appreciate your assistance. Thanks you.
Thank you for watching and for your kind comment. If all variables are stationary, you should have a stable model. However, given you are estimating a panel VAR model, it is important to understand the estimation procedure. Please follow the link below for a good survey of different methods of estimating a panel VAR model. For example, if you were to use a GMM procedure you should have larger N (number of countries) etc. Please check the links below. I hope it helps. Survey of Methods: www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1507.pdf?0ea674f009d6f2bffb1515c8b4b2cab6 GMM estimation in Stata paneldataconference2015.ceu.hu/Program/Michael-Abrigo.pdf
A very lucid explanation of such a tricky topic, sir! I have the following precise queries: 1) I have 2 variables with different order of integrations (I(2) & I(1)). Can I still use VECM model? If not, what could be the way out to model them? 2) Can we diagnose causality using VECM models? Thanking you!
No, you should not combine two series of I(2) and I(1). Depending on the context of your study, you may be able to use the first difference (or detrend your series) to make it stationary or at least of the same integration order. Granger causality tests are a set of F-tests which can be performed in bivariate or multivariate models. I hope that helps.
Sir, it seems like the second equation of the model specification in 7:58 is incorrect, because y2t-1 was written twice, please correct me if I'm wrong. Thank you sir for the best VAR lecture ever
A VAR model requires each individual variable to be I(0)? I thought that the stability of the system by itself would guarantee the stationarity of the VAR.
This was a great lecture. THANK YOU! I think the last 2 minutes were also critical. How VAR can be applied after transformation of the data (from I(1) to I(0)) but by doing that we are losing the benefits of VEC models (short-run vs long-run effects, if data, of course, fits the definitions). May I ask you a question. Are these takeaways change significantly with the exogenous variables? Can we also obtain IRFs in VAR and long-run/short-run effect estimations in VEC model for exogenous variables? Any source you would recommend? Thank you very much again. I could just finish this lecture today but I am planning to watch others as well.
Sir I have run VECM residuals diagnostic but my model found non normal and hetroskedastic residuals but it solution for it I already taking my variable as natural log form. What can I do for this problems Pls rpy
Sir, I have a doubt. Please help me. I have 8 independent variables and 32 data points. It's a time series model. Which methodology should I apply to find a long-term relationship, as I don't want to lose too many degrees of freedom?
Thanks for the lecture! May I clarify that all series have to be stationary before any lag selection and parameter estimation of the VAR model? Also, can we extend the concepts of Granger Causality and IRF to VARMA models?
Hello Sir, I want to estimate panel var through the estimate equation command to be able to include time and individual effects. But I do not know the coef covariance method to choose from the PANEL OPTIONS tab. My panel data has cross-section dependence. Can you help me with this?
Hello.. I'm Emma. I,m from Indonesia. Can I consultation about metode VAR with you? The formula that I use for my research is a little different from your explanation. Thanks.
، السلام عليكم دكتور شكرا على جهودك بوركت عندي طلب لو سمحت : I want to study the issue of the monetary policy transfer mechanism with the VARs form in the stata 16 program. Please help, send me your e-mail.
Hello Sir, I want to estimate panel var through the estimate equation command to be able to include time and individual effects. But I do not know the coef covariance method to choose from the PANEL OPTIONS tab. My panel data has cross-section dependence. Can you help me with this?
15:38 Choosing Lag Length
21:44 Reading VAR result
27:05 Granger Causality
35:41 Impulse
36:17 Variance Decomposition
38:19 Advantages of VAR Model
40:00 Problems with VAR
51:18 Why Stationarity Matters
1:09:08 Unit Root and Cointegration Tests
1:11:12 Engel-Granger Test
1:15:57 Granger Representation
1:20:56 VECM Johansen
this is the best explanation of VAR/VEC on youtube. Thanks
You explained in a very simple and objective manner a very difficult subject. Thank you for making this video!
Thank you for watching and for your comment and kind words. I am glad to know it helped.
This is the best econometrics video I have ever seen in my life !! Thank you for the best teaching method !! You should keep posting things more often !! You are doing a very very good job !!
I like the way you teach . It shows clearly what we have to do in steps. Thank you many times. MSc economics student , univ. of edinburgh.
THANK YOU is not enough to convey my gratitude to you.
This is an excellent video! it takes some confusing topics and presents them all in a really straightforward and logical manner. Many thanks!
Thank you for the class. You helped me a lot with this amazing explanation of a very difficult topic. I have read many books about the topic and now everything makes sense. I need to watch the video again, but I am pretty sure that, when finished, I will be ready to finish my last thesis chapter.
Thank you again. Thank you a lot.
Thank you so much for teaching this all in one topic. It made me understand it better. MSc Business (Finance Concentration) Canada
This is actually my first comment on UA-cam. I am currently writing my master's thesis in economics and have to learn VAR autodidactically - this is by far the best video I have found on UA-cam in this context. Hands-on and simply brilliant!
Thank you very much & keep on teaching like this! Greetings from Germany!
Thank you for your kind comments. I wish you all the best with your masters.
Beautiful mindful! Make life easy....keep it up Sir.
Thank you so much for following the lecture and for your kind comment.
Thank you so much for great explanation of VAR models. Best econometric video on UA-cam.
well curated.Thank u so much for the video😍
Thank you very much for this awesome lecture. Just a few questions:
1. 18:25 : What is "root of the matrix"? Eigenvalues? Also, what condition would I need to fulfil if I had a, say, VAR(3)?
2. 01:06:00 When you discussed ECM and argued about usability of the OLS, it was probably wise to mention, that the integrating factor is also I(0).
3. What about VECM models with more than 2 variables and higher dimensions (like, 3+ lags)? I guess we can get more than one integrating factor then? What about condition on the pi-matrix? Will there be just 1 pi-matrix or more than 1?
Up
For your question 1: I think the root of the matrix he mentioned are alpha and beta values of Yt-1 matrix, which is the coefficient when you fit the VAR(1) model. Similarly, when fitting VAR(3), there will be 2 additional A matrces (let's say A2 for lag 2 and A3 for lag 3). These matrices also need to be stable. The reason behind this is to ensure the variables in the model are stationary, which is consistent with the VAR model assumption (not too sure)
For question 2: I dont think using OLS necessarily mean the integrating factor is I(0). I think when he means is by differencing the 2 series, we now have 2 stationary series, whcih mean variance do not change over time and thus can use OLS.
VECM with more than 3 variables, lets say 3, the pi-matrix will be 3 by 3 matrix.
VECM with lags more than 1, let's say 2, there will be 2 pi-matrix.
Correct me if im wrong
It is really enlighting after watching your video, thanks a lot!
Thank you for watching. I am glad to hear it helped :-)
thank you, sir... you have a nice way of explaining each and everything in a topic .this is the very first lecture on youtube from which I learn the VAR method, you explained this topic in a very easy way and this lecture will help me in my research work. thanks again.
M.phil Student
university of Peshawar, Pakistan
Thank you for your comment. You may access the full course here hanomics.com/mnm038/
Thanks for posting your lecture!
You are welcome :-)
You made it very easy to grasp the difficult concepts.Amazing job
Thank you for watching and for your comment. I am glad to know it helped.
Great! you are a great teacher, i know one when i see one! :)
Excellent lecture with detailed and clear explanations!
Thanks for the video, very crisp and precise. Keep up the good work, very helpful indeed.
Thank you for your comment. You may access the full course here hanomics.com/mnm038/
It is a very useful lecture, thank you for sharing this video.
Thank you very much. Solving many problems in VAR and VECM.
Muito obrigado professor, ajudando um brasileiro alguns anos depois
Thank you for watching and your kind comments. I am glad to know it helps.
Agreed, the best concise, step by step presentation. Thank you.
Fabulous. Thank you!
Amazing lecture and explanation! Regards from Chile. Thanks!
بالتوفيق دائما دكتور
Thank you so much Dr. Hany! I really enjoyed the class.
it is the most wonderful lecture i have seen, thank you Professor:)
Thank you for watching and for your kind words.
Compliment, your lections are very interesting. You are a very good teacher :)
Thank you Sir, wonderful lecture
Thanku this video is very helpful for me. i want all lecture from this chennel.
Thanks for watching and for your kind comment. Glad to know it helped. Please subscribe to receive notifications for new content.
Thank you. This is excellent.
Thank you for your comment. I am glad to know it helped.
Hanomics
I have one question Sir
I have seven variables VAR Model, all endogenous. Should I do lag selection before stationerity or after stationerity at first difference...............my data is monthly data with 72 observations and I tried lag selection at first diff and AIC tells me 7 lags. The model does not become stable when I proceed for fcast and Irf analysis........kindly help me sir whether I shiuld select lags before I do stationerity at first diff.
Thank you in advance!
Thank You
Love this video.
Thank you for the lecture, it was very informative and helped me a lot.
Thank you so much for this wonderful lecture. I was really helpful in understanding VAR for the purpose of my thesis. I am currently doing a panel VAR analysis for 6 countries and have 6 variables. The problem I am experiencing is that, although all my variables are stationary at I(0) (using ADF, PP, LLC and IPS unit root tests), when I do the AR Roots test graph and table (using eviews 10) I find that my VAR is not stable regardless of the number of lags I use. What could be the problem and how can I fix it? I would like to interpret my Impulse Response Functions and Variance Decomposition. I would greatly appreciate your assistance. Thanks you.
Thank you for watching and for your kind comment. If all variables are stationary, you should have a stable model. However, given you are estimating a panel VAR model, it is important to understand the estimation procedure. Please follow the link below for a good survey of different methods of estimating a panel VAR model. For example, if you were to use a GMM procedure you should have larger N (number of countries) etc. Please check the links below. I hope it helps.
Survey of Methods: www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1507.pdf?0ea674f009d6f2bffb1515c8b4b2cab6
GMM estimation in Stata
paneldataconference2015.ceu.hu/Program/Michael-Abrigo.pdf
Hello Taimi, I ran into the exact same trouble as you did. I am wondering how did you resolve it? Would VECM be a solution?
This is a great lecture
Thank you for your comment. You may access the full course here hanomics.com/mnm038/
great work!
A very lucid explanation of such a tricky topic, sir!
I have the following precise queries:
1) I have 2 variables with different order of integrations (I(2) & I(1)). Can I still use VECM model? If not, what could be the way out to model them?
2) Can we diagnose causality using VECM models?
Thanking you!
No, you should not combine two series of I(2) and I(1). Depending on the context of your study, you may be able to use the first difference (or detrend your series) to make it stationary or at least of the same integration order. Granger causality tests are a set of F-tests which can be performed in bivariate or multivariate models. I hope that helps.
Great video! When are you posting Lecture 2?
Excellent Video
Sir, it seems like the second equation of the model specification in 7:58 is incorrect, because y2t-1 was written twice, please correct me if I'm wrong. Thank you sir for the best VAR lecture ever
Very helpful ......standard quality must applaud ......Sir plz can you suggest a book on basic time series econometrics
great teacher :)
Thank you for your comment :-)
Thank you so much. It is very helpful.
Thank you for the lecture
amazing lecture. Do you have any more regarding vecm , thanks
A VAR model requires each individual variable to be I(0)? I thought that the stability of the system by itself would guarantee the stationarity of the VAR.
Very helpful thank you.
Thank you very much
why the stability step is before the causality ?
why we test the stability of a model variables even do not causale
This was a great lecture. THANK YOU!
I think the last 2 minutes were also critical.
How VAR can be applied after transformation of the data (from I(1) to I(0)) but by doing that we are losing the benefits of VEC models (short-run vs long-run effects, if data, of course, fits the definitions).
May I ask you a question. Are these takeaways change significantly with the exogenous variables?
Can we also obtain IRFs in VAR and long-run/short-run effect estimations in VEC model for exogenous variables? Any source you would recommend?
Thank you very much again.
I could just finish this lecture today but I am planning to watch others as well.
Sir I have run VECM residuals diagnostic but my model found non normal and hetroskedastic residuals but it solution for it
I already taking my variable as natural log form.
What can I do for this problems
Pls rpy
Sir, I have a doubt. Please help me. I have 8 independent variables and 32 data points. It's a time series model. Which methodology should I apply to find a long-term relationship, as I don't want to lose too many degrees of freedom?
Thanks for the lecture! May I clarify that all series have to be stationary before any lag selection and parameter estimation of the VAR model? Also, can we extend the concepts of Granger Causality and IRF to VARMA models?
Dear professor,
Great video Just one question, in a VEC model, it is nor clear for me why the impulse response never converge to zero?
assalamalaikum, how i will get the PPT of your lecture.
In my macroeconomic data i have 10 + features How select the variabble in my dataset? Shall i Use correlation
I would usually rely on economic theory and/or existing literature
Sir what you do next if you found out the system is not stable?
Hello Sir, I want to estimate panel var through the estimate equation command to be able to include time and individual effects. But I
do not know the coef covariance method to choose from the PANEL OPTIONS tab. My panel data has cross-section dependence. Can you help me with this?
could you do one for var in stata
Can I use ECM for for an i(2) process if I use delta delta to make it stationary? and if so, does it change any of the terms I may include?
Good video, but way too many ads. Never seen these many ads pop up in other UA-cam videos.
Nothing to do with the video itself. Blame UA-cam/ Google.
I cannot still understand why it is called short-run dynamics
Hello..
I'm Emma. I,m from Indonesia. Can I consultation about metode VAR with you?
The formula that I use for my research is a little different from your explanation.
Thanks.
Is this a graduate level class?
Note for self.
1:00:02.
Nice lecture but having ads in every 4-5 minute breaks the concentration.
، السلام عليكم دكتور شكرا على جهودك بوركت عندي طلب لو سمحت : I want to study the issue of the monetary policy transfer mechanism with the VARs form in the stata 16 program. Please help, send me your e-mail.
Hello Sir, I want to estimate panel var through the estimate equation command to be able to include time and individual effects. But I
do not know the coef covariance method to choose from the PANEL OPTIONS tab. My panel data has cross-section dependence. Can you help me with this?