I am doing the same right now, thank you. Where can I find the video about log-transformed and level data? I want to use GDP, Inflation, FedRate, Loans to economy, and dont know which type of data to use - logs or levels. And also, how to transform levels into logs. Could you help please?
Hello Everyone! Thanks for watching! 🛎Buy Eviews Workfile Complete + SLIDES +Dataset (Includes the two VAR Videos material): payhip.com/b/V9ml ✅ Download the Dataset for FREE to replicate the resutls at: www.jdeconomics.com/vector-autoregression-models-in-eviews/ ✅The Tutorial is also available in STATA. Link: www.jdeconomics.com/how-to-estimate-var-models-in-stata/ ✅ Ensure to watch part 2: Watch PART 2: czcams.com/video/gfGsehtAmDw/video.html ✅ You can get access to all the EViews Workfiles, DO files (STATA) and Slides from my videos at: payhip.com/JDEconomics If you liked this video, please like and subscribe for more content! Your support helps me to create more video toturials. Please feel free to leave your comments and let me know if there is any topic you would like me to cover! ✅ Subscribe to my channel by clicking: czcams.com/channels/5P21WGFO4WRUlAiGLcwymg.html Thanks a lot! JD Economics.
Great video, one questio tho, on 3:37, does the coefficients before the variables have to take on the subscript of k, m, and p respectively? These coefficients do not have to be the same and are all estimated by OLS right?
Great explanation. I have a monthly data for 4 years. While estimating the optimal lag can i put 1:12 and then see the lag criteria table? Or must leave it as default 1:2 then look for the optimal lag? Coz as far as i understand that we can reduce the value of lags till we obtained the best model fit. Am i right?
Excellent explanation! i have a question, if the stability condition is satisfied, and there is no autocorrelation, but the causality tests shows, there isnt granger causality, it means that we can not use this model?
Hi Javier, not necessarily. Are the IRF sensible? Do they have an economic interpretation that makes sense? Those are some of the questions you need to go through when doing the analysis. Regards, JD.
@@JDEconomics Hi! I ran a standard VAR model and the estimation shows that only a few coefficients are significant, also there is no granger causality. However, literature says that my X variable has an impact on Y variable. 1. in absence of granger causality, can i still run VAR? 2. Does the VAR make sense even if coefficients are insignificant (very low t Value)? 3. does the IRF make sense in a standard VAR? is it not necessary to introduce restrictions due to possibility of correlation inn error terms? Thanks in advance!
@@vijayaagrawal7004 hi! Hard to comment because I don’t have the model to see it. Feel free to send me an email. You can contact me through my website. Www.JDEconomics.com Thanks!
Hello sir, thank you for your explanation, please you make explanation about VECM in panel data because i need to learn about VECM panel data to my task in school 🙏🏻 Thankyou 🙏🏻
Is there a way to customize the shock size without add-ins? I have a trivariate VAR (GDP, gov. spending and tax revenue) and want to apply a spending shock on GDP to estimate fiscal multipliers.
Hi sir, i have a question , For the number of lags selected in my case based on the criteria (LR, FPE, AIC) is 8, and according to (SC, HQ) lag=1, so I chose to use lag 8. I performed the estimation and tested the significance of the coefficients. I also tested the stability of the model, and the results show that no root is outside the unit circle. However, when I performed the residual autocorrelation test with the chosen lag of 8, I found that the probability for lag 8 is 0.000, which is less than 0.05, indicating autocorrelation. In this case, should I change the number of lags? Can I choose a lower lag, such as 3, where there is no autocorrelation? If lags 3, 4, 6, and 7 are not autocorrelated, which one should I choose among them,Also, is the absence of autocorrelation the only condition I need to verify, considering that the residuals do not follow a normal distribution in my case? Thanks😊
Dear Sir, I tried to replicate the results of Stock and Watson (2001, JEP) using quarterly data from 1960:q1 to 2023:q4, but I found that while inflation and unemployment rates are stationary, but the Federal funds rate is non-stationary. Is it okay to for them to simply use the original series of the Federal funds rate to construct the VAR model?
Hi, Thanks for your email. It's always a good practice to search in google scholar/NBER or other academic sites for the topic and see what is the existing literature. That way you know what their methodology approach is and what are the existing questions about the topic. You can email me at jdeconomics.inquiries@gmail.com if needed. Regards,
@@bellisma77I have replicated the model the way the authors estimated it. I believe the variables are stationary at the 10% significance level. I can’t rememver as I uploaded this video over 2 years ago.
I really can not describe how much your videos helped me. My question is in general not about the video 😊. If my variables are I(1), should i have to use variables at levels when estimating VAR or VECM ? Thank you sooo much
Hello, Thank you for your message. I have estimated the model respecting the original paper. Variables in levels and 4 lags. It's a replica. You can test the stationarity of the variables. I can't remember at the time if they were all stationary at least at the 10% level. Regarding Cointegration, you would check for cointegration in the case your variables are non stationary but they have some long run relationship. If the variables are cointegrated they keep a long run relationship. In such case, you need to estimate the VAR model using a vector error correction model(VECM) which you can select in Eviews when estimating a VAR. I plan to upload a video explaining the steps for cointegration. Regards, JD
Good evening sir, can you please provide me any reference where the lag specification is mentioned that for annual data 1 lag is sufficient , for quarterly:12 lags, etc. Thankyou
Hi Kajal. Thank you for your message. I can definitely look around my books and find a reference for this matter. However, I didn't say it's sufficient. I mentioned as a note that there should be no autocorrelation at the lag selected regardless of using an information criteria or using the data frequency as a guide. I can definitely suggest you to read a paper by Lutz Kilian and Ivanov, where they cover which information criteria is better (AIC, BIC, HQ, etc..) depending on the frequency of data. For example: They find HQ is better for quarterly data. Hope it helps! Here is the link: drphilipshaw.com/Protected/A%20Practitioners%20Guide%20to%20Lag%20Order%20Selection%20for%20VAR%20Impulse%20Response%20Analysis.pdf
@@JDEconomics Ok, thankyou so much sir for your help. Your videos are really informative. I request you to please make a video on ARDL model, that will be really helpful. Thankyou again :)
Hi, I'm creating a VAR model that has 3 endogenous variables- turnover, market return and individual return, and an exogenous variable, volatility. I would like to find the lag for the exogenous variable which according to the literature is not the same as the endogenous variables. How do I go about finding the lag structure for my exogenous variable?
Hello sir please what can cause the THRESHOLD technic to disappear from the available technique in e-vews if one want to run a data using the threshold analysis for a number of selected countries in the ecowas region.
Thank you so much for your understandable video, but in case of my model I still have doubt, could you please help me sir? I have used lag lenght criteria, and changed lags in any kinds, I still have autocorrelation on the model, how could I fix that serial correlation? How can I improve the model ? 😟
Hi Buyankhishig, Thank you for your comment. I would need to see the details, but is there any big event (i.e., 1970, 2000, 2008)? Sometimes you can include a dummy variable ,say for the year 2008 as an exogenous variable (you create a dummy =1 for the year where there is a sharp autocorrelation and add it in the model in the box that says "exogenous variables"). Regards, JD Econ.
Great video. Very helpful and informative. I'd be more appreciate if you can interpret the results in more detail. For example, at 11:00 could you explain the difference between "No serial correlation at lag h " and "No serial correlation at lags 1 to h"?
Thanks! I can only do so much in the video. I tried to cover ad many things I could. I taught you guys how to walk. Now you are in a good spot to try on your own and learn how to run. The idea of the videos is that you guys complement it with reading. Good luck!
Thank you for your wonderful lecture! I got a question that should we care about the unit root test? Since I think VAR is applicable only if the variables are stationary after taking 1st difference, could you talk about that a little bit? (e.g When should apply unit root test and cointegration test) Thank you!
Hello will, Thanks for your message! You are on the right track. I have respected the author's Stock and Watson 2001 paper and have used the variables in levels as they did. The dataset I have is not the original used by the authors, so there may be a minor discrepancy in the stationarity tests, however, you should use for your research paper variables in their stationary form. I hope that helped to clarify your question. I have another tutorial where I teach unit root tests, and I think if you are a Stata user I will be submitting sometime next week a step by step tutorial with the unit root tests as well. In case you want that material, it is ready to buy for 3.99, the DO set +slides +dataset. payhip.com/b/cq9LQ Kind Regards, JD
Hi, this is a great explanation on VAR with long run restriction, very easy to understand. much thanks :) I have a question, should we use VECM if our variables are cointegrated? Or is there any other way to use SVAR model with cointegration? I try to understand the model used in Dungey and Vehbi's "The influences of international output shocks from the US and China on ASEAN economies" but get lost. Thank you very much for your response :)
Thank you sir, your explanation really help me on my thesis. I have question regarding to Stability test, do you have any recommendation in case there are several points that have value > 1? Is it still a problem eventhough the value is not very far from 1 i.e. 1.003527?
Those cases are very particular and depend on the researcher what argument you make to determine whether it is valid or not. I would suggest to try with a different amount of lags from one of the model selection criterias (AIC, HQ, SW) Good luck with your thesis! Feel free to subscribe for more tutorials! Regards, JD
Hi Is it possible to select different lag lengths for each of the variables? e.g. I want the inflation and fed-rate to have 2 but the unemployment to have 4? Is that possible in EViews? And how would that work if I want to estimate a mixed frequency VAR? Thank you!
Hi Layla, Thanks for your message. You can specify the lags using intervals. If you see by default is 1 2, you can keep typing more lags in intervals. i.e, 1 2 4 6 8 9. Etc. For more details check in section (Estimating VAR): www.eviews.com/help/helpintro.html#page/content/VAR-Vector_Autoregressions_(VARs).html Regarding mixed frequency vars, its only avaialble I believe starting on Eviews 11 and here is a link with the details: www.eviews.com/EViews11/ev11ecest_n.html They have a video on how to estimate it. Regards, JD
Thank you very much for the tutorial.. but I have a question regarding last minutes of your video about granger casuality test.. in the table.. it shows under unemployment and fed rate.. there "All" whose significant is 0.0003 which is less than 0.05, so can we say that unemployment and fed rate simultaneously ganger causes inflation?? Cause I think that "All" represents unemployment and fed rate variables..or any other interpretations? thanks
Hi, No worries! Thanks for your comment. Regarding your question, the statistic in the last row (All) is the chi square-statistic for joint significance. Unemployment and the Fed Rate together help to predict inflation, but the Fed Rate on its own does not. I hope that helps to answer your question. Best Regards, JD.
Thanks a lot Juan! But just one question, weren't you supposed to use differences or log transformations to make your variables stationary before running the regression?
Hi can u answer this? If I'm using a var model to estimate three varibles where is an exogenous varible according to economic theory eg. Crude oil prices. The other varibles im using is money supply and exchange. Does the equations look the same or it changes cuz of the oil price varibles.
Hello Lisa, Thank you for your message. I have actually worked in my paper with crude oil prices. I strongly suggest you to read the paper by Lutz Killian 2009 - (not all shocks are alike...). You need to use a structural identification, a lower triangular matrix like the one I show in the second video of VAR models. Here is the link for that paper of Lutz Killian: jpkc.whut.edu.cn/ppkc1/gjjjx/images_1/2013103036759173.pdf. Also, you may want to email me directly at the mail in the description so I can clear any doubts if you need. In the exogenous variables just a constant or a dummy if the residuals on your VAR are not normally distributed.
Hi, the variables in the var model should be stationary. For the model I estimated, I have replicated the paper by S&W2001, which has used them in levels. Check my tutorial in Stata for VAR models where I estimate a model in differences. Regards, JD
First, thank you so much for these tutorials and informative videos. I just have a very basic question. When we proceed with VAR model, is it possible to include control variables as well? Or we are just more interested in the relationship within variables of interest? Thank you...
Hello, thạnk you so much for your explanation, it is soooo helpful. If possible, could you please clarify these points for me? It would mean a lot! 1. Is it compulsory that all the variables must be stationary at first difference for us to apply VAR model? If they are a mixture of both I(0) and I(1) then we should use ARDL? 2. If the variables are stationary at first difference, then when we run the VAR estimates and IRF and Variance decomposition, we use their difference or their level/log form?
Hi, Thanks for your message. If the variables are non stationary, my suggestion is to check for cointegration. If they are cointegrated, then you can estimate a VECM. In case they are not stationary-not cointegrated, then use differences and the model should be used in differences.
@@JDEconomics Thank you soo much for your response. As seen in many papers, after doing the unit root test and confirming that the variables are stationary at first difference, many authors continue with their impulse response function and variance decomposition with the variables in level form? Does that make any sense? Because I thought if the variables are I(1), then the IRF and VD should be carried out in the difference form? Thank you!
@@anhnguyentu3531 The shocks have permanent effects if they are non stationary .Use it in differences. Maybe the papers you saw use vec and not var. Send me some samples
It can have to do with stationarity of the variables you are using, or also, it can be that there is strong autocorrelation in a residual. You can check the residual series, and add a dummy variable as exogenous variable in the VAR. Regards, JD
@@JDEconomics That'd be great, thanks for the quick reply! Your explanation is great and I'd love to replicate the results on my eviews. At the moment, Google drive says that I'm not authorised to access the file
If when doing the unit root test for 5 variables, the result shows 3 variables are stationer and the other 2 is not-stationer, are we supposed to do first differencing to all the variables (Regardless the variable already stationer at level or not) and then test for cointegration to decide to use VAR or VECM ? Or should we use ARDL ?
Hey you can use an ARDL model. I don't recommend forcing transformations that are not necessary (i.e., using 1st differences on stationary variables). Also, Make sure you check with the diverse unit root tests. I have a tutorial about it. Good luck!
@@diazjubairy1729 sounds good! Feel free to check my website and register to the newsletter (I don’t send emails all day. Don’t worry). I will just notify when a new video comes out and free webinars too, where I go live and go through some examples and questions. Take care!
The thing that i didn't understand is you haven't checked the situation if the inflation or the other variables are at their stationary level or not, why you havent
Thank you for your explanation! Could you tell me what is the equation about the residual series test in 11:10? I think there would be the three equations for each dependent variables.. And please let me know where can I check the term definition of each table in the Eview!
Hi sir, in your example there are 2 dependent variables while you do the VAR analysis. However, if I have only one dependent variable & more than 1 independent variables in my research, can I still use VAR as my statistical method?
Hi Omer, thank you for your message. There is a second video that will be posted during the weekend where I cover IRFS and variance decomposition. Let me know if there is anything you need help with. Cheers
You, my good sir, have just saved my Master's thesis. Thank you very much.
Excellent! I am glad to hear so. Cheers! JD
I am doing the same right now, thank you. Where can I find the video about log-transformed and level data? I want to use GDP, Inflation, FedRate, Loans to economy, and dont know which type of data to use - logs or levels. And also, how to transform levels into logs. Could you help please?
@@JDEconomics may i have your email id please?
@@md.litonali1425 Jdeconomics.inquiries@gmail.com , also, all my info is in my website. Regards,
thanks a lot , it is very clear and easy to apply
You are most welcome
Hello Everyone! Thanks for watching!
🛎Buy Eviews Workfile Complete + SLIDES +Dataset (Includes the two VAR Videos material): payhip.com/b/V9ml
✅ Download the Dataset for FREE to replicate the resutls at:
www.jdeconomics.com/vector-autoregression-models-in-eviews/
✅The Tutorial is also available in STATA. Link: www.jdeconomics.com/how-to-estimate-var-models-in-stata/
✅ Ensure to watch part 2: Watch PART 2: czcams.com/video/gfGsehtAmDw/video.html
✅ You can get access to all the EViews Workfiles, DO files (STATA) and Slides from my videos at: payhip.com/JDEconomics
If you liked this video, please like and subscribe for more content! Your support helps me to create more video toturials. Please feel free to leave your comments and let me know if there is any topic you would like me to cover!
✅ Subscribe to my channel by clicking:
czcams.com/channels/5P21WGFO4WRUlAiGLcwymg.html
Thanks a lot!
JD Economics.
thx man
Ur welcome! Regards, JD
Thank you very much for doing this content and help students . Perfect explanation in a very easy language!
Thanks Renata! Good Luck!
Very good explanation. Thank you.
Thabks Ibrahim! I will be submitting more tutorials shortly. Feel free to subscribe to get notified. Good luck! Regards, JD
this is super useful, Thank you very much Sir
You’re welcome!
Wow!! Great,thank you, very precise
No worries! Glad you liked it. Kind Regards, JD
Good job man.
Thanks! make sure to watch part 2, and feel free to subscribe if you are interested in more content. Take care! Good luck
Great video, one questio tho, on 3:37, does the coefficients before the variables have to take on the subscript of k, m, and p respectively? These coefficients do not have to be the same and are all estimated by OLS right?
Thank you and vamos Columbia ❤️
Thanks! please make sure to check my website where you will find all the material I have available!
LINK: www.jdeconomics.com
Best Regards,
JD
Great explanation. I have a monthly data for 4 years. While estimating the optimal lag can i put 1:12 and then see the lag criteria table? Or must leave it as default 1:2 then look for the optimal lag? Coz as far as i understand that we can reduce the value of lags till we obtained the best model fit. Am i right?
Hi! Yes! You can estimate with 12 lags and then adjust it based on the lag length criteria. Regards, JD
Thanks for the explanation. Question: Is there a rule to choose the lags to make the Lm Autocorrelation for any VAR Model?
Not really. But should be somewhere close to the amounts of lags you are using in the model. Regards, JD.
thx for that great video !
No problem! Thank you for your feedback. Good luck! JD
Excellent explanation! i have a question, if the stability condition is satisfied, and there is no autocorrelation, but the causality tests shows, there isnt granger causality, it means that we can not use this model?
Hi Javier, not necessarily. Are the IRF sensible? Do they have an economic interpretation that makes sense? Those are some of the questions you need to go through when doing the analysis. Regards, JD.
Yes they do have an economic interpretation and IRF are sensible! Thanks a lot!!!
@@javiermoreno701 great! Feel free to subscribe (if you haven’t) for more videos coming! Good luck! JD
@@JDEconomics Hi! I ran a standard VAR model and the estimation shows that only a few coefficients are significant, also there is no granger causality. However, literature says that my X variable has an impact on Y variable.
1. in absence of granger causality, can i still run VAR?
2. Does the VAR make sense even if coefficients are insignificant (very low t Value)?
3. does the IRF make sense in a standard VAR? is it not necessary to introduce restrictions due to possibility of correlation inn error terms? Thanks in advance!
@@vijayaagrawal7004 hi! Hard to comment because I don’t have the model to see it. Feel free to send me an email. You can contact me through my website. Www.JDEconomics.com
Thanks!
Hello sir, thank you for your explanation, please you make explanation about VECM in panel data because i need to learn about VECM panel data to my task in school 🙏🏻
Thankyou 🙏🏻
Thanks for your comment.! I will add it to the list for the future. Regards, JD
thank you sir
You are great!
Thanks!
great lesson Sir.. But your website is down so not able to download the dataset
Thank! I have updated the link. Let me know if now it works. Cheers
Is there a way to customize the shock size without add-ins? I have a trivariate VAR (GDP, gov. spending and tax revenue) and want to apply a spending shock on GDP to estimate fiscal multipliers.
Hi JD, thank you so much for this helpful video. One question, after estimating the VAR model, do you have a video on how to forecast the estimates?
Thanks! Sorry, no
Hi sir, i have a question , For the number of lags selected in my case based on the criteria (LR, FPE, AIC) is 8, and according to (SC, HQ) lag=1, so I chose to use lag 8. I performed the estimation and tested the significance of the coefficients. I also tested the stability of the model, and the results show that no root is outside the unit circle. However, when I performed the residual autocorrelation test with the chosen lag of 8, I found that the probability for lag 8 is 0.000, which is less than 0.05, indicating autocorrelation.
In this case, should I change the number of lags? Can I choose a lower lag, such as 3, where there is no autocorrelation? If lags 3, 4, 6, and 7 are not autocorrelated, which one should I choose among them,Also, is the absence of autocorrelation the only condition I need to verify, considering that the residuals do not follow a normal distribution in my case? Thanks😊
What is difference b/w VAR model and VAR estimates?? and also deference b/w test and model in research.
Dear Sir, I tried to replicate the results of Stock and Watson (2001, JEP) using quarterly data from 1960:q1 to 2023:q4, but I found that while inflation and unemployment rates are stationary, but the Federal funds rate is non-stationary. Is it okay to for them to simply use the original series of the Federal funds rate to construct the VAR model?
sir can you please help for checking impact of Public private partnership on GDP of any country thank you
Hi, Thanks for your email. It's always a good practice to search in google scholar/NBER or other academic sites for the topic and see what is the existing literature. That way you know what their methodology approach is and what are the existing questions about the topic. You can email me at jdeconomics.inquiries@gmail.com if needed.
Regards,
What about unit root test? If my variables are I(1) should we must used the first differened variables when estimating VAR?
Hi! If your variables are non-stationary, your variables need to be transformed into stationary.
Thanx for replying. What i noticed that for estimating VAR model you used variables at their levels.
@@bellisma77I have replicated the model the way the authors estimated it. I believe the variables are stationary at the 10% significance level. I can’t rememver as I uploaded this video over 2 years ago.
I really can not describe how much your videos helped me.
My question is in general not about the video 😊. If my variables are I(1), should i have to use variables at levels when estimating VAR or VECM ? Thank you sooo much
@@bellisma77for a var in I(1)
Sir, you haven't shown the stationarity of the variables used. Are they all stationary at level? Also when do we use cointegration?
Hello, Thank you for your message. I have estimated the model respecting the original paper. Variables in levels and 4 lags. It's a replica. You can test the stationarity of the variables. I can't remember at the time if they were all stationary at least at the 10% level. Regarding Cointegration, you would check for cointegration in the case your variables are non stationary but they have some long run relationship. If the variables are cointegrated they keep a long run relationship. In such case, you need to estimate the VAR model using a vector error correction model(VECM) which you can select in Eviews when estimating a VAR. I plan to upload a video explaining the steps for cointegration. Regards, JD
Good evening sir, can you please provide me any reference where the lag specification is mentioned that for annual data 1 lag is sufficient , for quarterly:12 lags, etc. Thankyou
Hi Kajal. Thank you for your message. I can definitely look around my books and find a reference for this matter. However, I didn't say it's sufficient. I mentioned as a note that there should be no autocorrelation at the lag selected regardless of using an information criteria or using the data frequency as a guide. I can definitely suggest you to read a paper by Lutz Kilian and Ivanov, where they cover which information criteria is better (AIC, BIC, HQ, etc..) depending on the frequency of data. For example: They find HQ is better for quarterly data. Hope it helps!
Here is the link: drphilipshaw.com/Protected/A%20Practitioners%20Guide%20to%20Lag%20Order%20Selection%20for%20VAR%20Impulse%20Response%20Analysis.pdf
@@JDEconomics Ok, thankyou so much sir for your help. Your videos are really informative. I request you to please make a video on ARDL model, that will be really helpful. Thankyou again :)
Hi, I'm creating a VAR model that has 3 endogenous variables- turnover, market return and individual return, and an exogenous variable, volatility. I would like to find the lag for the exogenous variable which according to the literature is not the same as the endogenous variables. How do I go about finding the lag structure for my exogenous variable?
the goat 100%
Thanks!
Hello sir please what can cause the THRESHOLD technic to disappear from the available technique in e-vews if one want to run a data using the threshold analysis for a number of selected countries in the ecowas region.
Thank you so much for your understandable video, but in case of my model I still have doubt, could you please help me sir? I have used lag lenght criteria, and changed lags in any kinds, I still have autocorrelation on the model, how could I fix that serial correlation? How can I improve the model ? 😟
Is there any approach I can get some help? 🙏
Hi Buyankhishig, Thank you for your comment. I would need to see the details, but is there any big event (i.e., 1970, 2000, 2008)? Sometimes you can include a dummy variable ,say for the year 2008 as an exogenous variable (you create a dummy =1 for the year where there is a sharp autocorrelation and add it in the model in the box that says "exogenous variables"). Regards,
JD Econ.
Great video. Very helpful and informative. I'd be more appreciate if you can interpret the results in more detail. For example, at 11:00 could you explain the difference between "No serial correlation at lag h " and "No serial correlation at lags 1 to h"?
Thanks! I can only do so much in the video. I tried to cover ad many things I could. I taught you guys how to walk. Now you are in a good spot to try on your own and learn how to run. The idea of the videos is that you guys complement it with reading. Good luck!
Thank you for your wonderful lecture! I got a question that should we care about the unit root test? Since I think VAR is applicable only if the variables are stationary after taking 1st difference, could you talk about that a little bit? (e.g When should apply unit root test and cointegration test) Thank you!
Hello will, Thanks for your message! You are on the right track. I have respected the author's Stock and Watson 2001 paper and have used the variables in levels as they did. The dataset I have is not the original used by the authors, so there may be a minor discrepancy in the stationarity tests, however, you should use for your research paper variables in their stationary form. I hope that helped to clarify your question. I have another tutorial where I teach unit root tests, and I think if you are a Stata user I will be submitting sometime next week a step by step tutorial with the unit root tests as well. In case you want that material, it is ready to buy for 3.99, the DO set +slides +dataset.
payhip.com/b/cq9LQ
Kind Regards,
JD
@@JDEconomics Thank you for your wonderful reply!
@@willwu5366 No problem! good luck!
Hi, this is a great explanation on VAR with long run restriction, very easy to understand. much thanks :)
I have a question, should we use VECM if our variables are cointegrated? Or is there any other way to use SVAR model with cointegration? I try to understand the model used in Dungey and Vehbi's "The influences of international output shocks from the US and China on ASEAN economies" but get lost.
Thank you very much for your response :)
Is it not necessary to check the stationarity of variables in var
Hi. Variables should be stationary. In this tutorial I just replicate the paper from Stock and Watson for instructional purposes. Cheers
Thank you sir, your explanation really help me on my thesis.
I have question regarding to Stability test, do you have any recommendation in case there are several points that have value > 1? Is it still a problem eventhough the value is not very far from 1 i.e. 1.003527?
Those cases are very particular and depend on the researcher what argument you make to determine whether it is valid or not. I would suggest to try with a different amount of lags from one of the model selection criterias (AIC, HQ, SW)
Good luck with your thesis! Feel free to subscribe for more tutorials!
Regards,
JD
Hi Is it possible to select different lag lengths for each of the variables? e.g. I want the inflation and fed-rate to have 2 but the unemployment to have 4? Is that possible in EViews? And how would that work if I want to estimate a mixed frequency VAR? Thank you!
Hi Layla, Thanks for your message. You can specify the lags using intervals. If you see by default is 1 2, you can keep typing more lags in intervals. i.e, 1 2 4 6 8 9. Etc. For more details check in section (Estimating VAR): www.eviews.com/help/helpintro.html#page/content/VAR-Vector_Autoregressions_(VARs).html
Regarding mixed frequency vars, its only avaialble I believe starting on Eviews 11 and here is a link with the details: www.eviews.com/EViews11/ev11ecest_n.html
They have a video on how to estimate it.
Regards,
JD
Thank you very much for the tutorial.. but I have a question regarding last minutes of your video about granger casuality test.. in the table.. it shows under unemployment and fed rate.. there "All" whose significant is 0.0003 which is less than 0.05, so can we say that unemployment and fed rate simultaneously ganger causes inflation?? Cause I think that "All" represents unemployment and fed rate variables..or any other interpretations? thanks
Hi, No worries! Thanks for your comment. Regarding your question, the statistic in the last row (All) is the chi square-statistic for joint significance. Unemployment and the Fed Rate together help to predict inflation, but the Fed Rate on its own does not. I hope that helps to answer your question. Best Regards, JD.
@@JDEconomics that helps a lot really 🙏👍👍 it answers my question
@@agungwangsa7453 My pleasure. Take care. JD
@@JDEconomics by the way.. we are looking forward to Cointegration and VECM test.. 🙏
Thanks a lot Juan! But just one question, weren't you supposed to use differences or log transformations to make your variables stationary before running the regression?
Hi, yes. I just replicated the paper as per the authors. Regards
@@JDEconomics Understood. Thanks Juan! Shalom :)
Thank you sir....
Sir can you please answer my questions that electronic payments concept is worthless in developed countries.
Ur welcome!
Thanks a lot!!!
No worries! Check my website www.jdeconomics.com
Cheers!
can you use dummy variables as exogenous variables in var and vecm? thankyou
You can. There is an exogenous variable section in the var window. Regards, JD
thank you very much, thats really helpful
Hi can u answer this?
If I'm using a var model to estimate three varibles where is an exogenous varible according to economic theory eg. Crude oil prices. The other varibles im using is money supply and exchange. Does the equations look the same or it changes cuz of the oil price varibles.
Hello Lisa, Thank you for your message. I have actually worked in my paper with crude oil prices. I strongly suggest you to read the paper by Lutz Killian 2009 - (not all shocks are alike...). You need to use a structural identification, a lower triangular matrix like the one I show in the second video of VAR models. Here is the link for that paper of Lutz Killian: jpkc.whut.edu.cn/ppkc1/gjjjx/images_1/2013103036759173.pdf.
Also, you may want to email me directly at the mail in the description so I can clear any doubts if you need. In the exogenous variables just a constant or a dummy if the residuals on your VAR are not normally distributed.
I would email and thanks for responding😊
Isn't it necessary that variables should be in stationary?
If the series is non-stationary do we include the first difference variables or at the level form.
Hi, the variables in the var model should be stationary. For the model I estimated, I have replicated the paper by S&W2001, which has used them in levels. Check my tutorial in Stata for VAR models where I estimate a model in differences. Regards, JD
First, thank you so much for these tutorials and informative videos. I just have a very basic question. When we proceed with VAR model, is it possible to include control variables as well? Or we are just more interested in the relationship within variables of interest? Thank you...
You can add them as exogenous variables. Regards
@@JDEconomics ¡Muchas gracias! Thank you so much ^__^
Hello, thạnk you so much for your explanation, it is soooo helpful. If possible, could you please clarify these points for me? It would mean a lot!
1. Is it compulsory that all the variables must be stationary at first difference for us to apply VAR model? If they are a mixture of both I(0) and I(1) then we should use ARDL?
2. If the variables are stationary at first difference, then when we run the VAR estimates and IRF and Variance decomposition, we use their difference or their level/log form?
Hi, Thanks for your message. If the variables are non stationary, my suggestion is to check for cointegration. If they are cointegrated, then you can estimate a VECM. In case they are not stationary-not cointegrated, then use differences and the model should be used in differences.
@@JDEconomics Thank you soo much for your response. As seen in many papers, after doing the unit root test and confirming that the variables are stationary at first difference, many authors continue with their impulse response function and variance decomposition with the variables in level form? Does that make any sense? Because I thought if the variables are I(1), then the IRF and VD should be carried out in the difference form?
Thank you!
@@anhnguyentu3531 The shocks have permanent effects if they are non stationary .Use it in differences. Maybe the papers you saw use vec and not var. Send me some samples
If VAR does not satisfy the stability condition the what will be the solution ? please suggest
It can have to do with stationarity of the variables you are using, or also, it can be that there is strong autocorrelation in a residual. You can check the residual series, and add a dummy variable as exogenous variable in the VAR.
Regards, JD
Dear all. Can you help me to estimate threshold var model in r logiciel.
The link to download the free dataset doesn't work...
Thanks. I’ll fix it today. Regards
@@JDEconomics That'd be great, thanks for the quick reply! Your explanation is great and I'd love to replicate the results on my eviews. At the moment, Google drive says that I'm not authorised to access the file
If when doing the unit root test for 5 variables, the result shows 3 variables are stationer and the other 2 is not-stationer, are we supposed to do first differencing to all the variables (Regardless the variable already stationer at level or not) and then test for cointegration to decide to use VAR or VECM ?
Or should we use ARDL ?
Hey you can use an ARDL model. I don't recommend forcing transformations that are not necessary (i.e., using 1st differences on stationary variables). Also, Make sure you check with the diverse unit root tests. I have a tutorial about it. Good luck!
@@JDEconomics thank you for the answer. Will look into your video
@@diazjubairy1729 sounds good! Feel free to check my website and register to the newsletter (I don’t send emails all day. Don’t worry). I will just notify when a new video comes out and free webinars too, where I go live and go through some examples and questions. Take care!
The thing that i didn't understand is you haven't checked the situation if the inflation or the other variables are at their stationary level or not, why you havent
Hi. I have replicated the original paper the way it is. For accurate estimation and forecasting, variables should be stationary. Regards
Thank you for your explanation!
Could you tell me what is the equation about the residual series test in 11:10?
I think there would be the three equations for each dependent variables..
And please let me know where can I check the term definition of each table in the Eview!
Hi sir, in your example there are 2 dependent variables while you do the VAR analysis. However, if I have only one dependent variable & more than 1 independent variables in my research, can I still use VAR as my statistical method?
Hi, Variables in VAR models are endogenous, Kind Regards. JD
I could not any interpretation about VAR results.
Hi Omer, thank you for your message. There is a second video that will be posted during the weekend where I cover IRFS and variance decomposition. Let me know if there is anything you need help with. Cheers
Thank you
No problem !!