Vector Error Correction Model (VECM) - Step 4 of 4

แชร์
ฝัง
  • เผยแพร่เมื่อ 1 ธ.ค. 2024

ความคิดเห็น • 101

  • @ImperiumLearning
    @ImperiumLearning 5 ปีที่แล้ว +11

    This 4-part series was really useful for me when I was studying for my MSc. So much so, that it inspired me to upload 5 of my own videos on estimating a VECM and how to interpret the results.

  • @samueljlaltlanzaua363
    @samueljlaltlanzaua363 5 ปีที่แล้ว +5

    Thanks so much, Professor. I had searched for so long time the kind of your video. All the best. From Mizoram, India

  • @okunlolaacademy8682
    @okunlolaacademy8682 3 ปีที่แล้ว

    You are one of my best mentors. Thank you

  • @onuongamoraa9690
    @onuongamoraa9690 5 ปีที่แล้ว +4

    Your explanations are very clear.Thanks for this

  • @pitchayadasrivipath6772
    @pitchayadasrivipath6772 6 ปีที่แล้ว +2

    thank you so much!! you have brought me to another level of understanding of VECM. :)

  • @pramodpandey8316
    @pramodpandey8316 5 ปีที่แล้ว +3

    Sir thank you very much for such a simple and nice interpretation. its a great help to the researchers.

  • @MrHugosky1
    @MrHugosky1 6 ปีที่แล้ว +3

    Magnificent presentation! Thank you very much for the lesson!

  • @polapallyvenugopal656
    @polapallyvenugopal656 6 ปีที่แล้ว

    thank you pat for helping people in understanding VECM model.

  • @abdullahjurat1956
    @abdullahjurat1956 7 ปีที่แล้ว +2

    You got a great teaching skill. Keep it up!

  • @lucabutiniello568
    @lucabutiniello568 4 ปีที่แล้ว +1

    Crystal clear as always, thank you very much prof!

  • @fionncliffe5706
    @fionncliffe5706 3 ปีที่แล้ว +1

    Thanks Pat, excellent insight.

  • @milktea2021
    @milktea2021 5 ปีที่แล้ว +2

    This is very helpful Sir. Thank you so much for sharing this materias. 🙏🏻

  • @emilierutledge3933
    @emilierutledge3933 7 ปีที่แล้ว +3

    Thanks so much Pat, your videos are very informative indeed! Can you please do a video to show how you can impose a cointegrating restriction using Eviews? I couldn't find a video on this anywhere....

  • @Redtri7kala
    @Redtri7kala 5 ปีที่แล้ว +1

    sir can you help me? i have panel data and i make the steps you say but in the end on eviews it doesent appears the lm test or cusum test?why ??please help me

  • @nausheensodhi803
    @nausheensodhi803 2 ปีที่แล้ว

    Thank you for this video! However, I am still not clear if I can use VECM for panel data containing 4 variables, 2 of which are I(1) and 2 are I(0). Could you please tell me if VECM can be used? Thanks!

  • @weena1725
    @weena1725 ปีที่แล้ว

    Question: The VEC model have two cointegrating equations. The coefficient, C(1) is negative and significant, but C(2) is positive and significant, what will be the overall conclusion for longrun equilibrium?

  • @SagangaKapaya
    @SagangaKapaya 5 ปีที่แล้ว +2

    Perfect explanations there, thanks

  • @melisgultekin9881
    @melisgultekin9881 3 ปีที่แล้ว +1

    Thanks so much for your explanatory video. I found 2 vectors with my variables with the helpof Johansen test so I want to ask what if my 2nd cointegration equaiton CE(2) is positive and not significant, I can still say that there is long run relationship? because my 1st cointegrated equation CE(1) is negative and significant. Please help me about this issue. Thanks..

  • @mickeykozzi
    @mickeykozzi 4 ปีที่แล้ว +1

    Hello Dr. Firstly thank you for your videos.
    I have a questions regarding your residual testing.
    It appears that you are using the chi-squared P-value rather than the F-stat p-value within your residual testing (LM, hetro).
    Why do you choose chi squared over f-stat?
    Thank you Dr :)

    • @PatObi
      @PatObi  4 ปีที่แล้ว +1

      I believe any can be used.

    • @mickeykozzi
      @mickeykozzi 4 ปีที่แล้ว +1

      @@PatObi thank you

  • @irinaalexandrageorgescu6865
    @irinaalexandrageorgescu6865 4 ปีที่แล้ว

    Very helpful! Congratulations!

  • @AbstrakVoice
    @AbstrakVoice 7 ปีที่แล้ว +2

    Dear sir, can u make other example about VECM analysis? i am very insterest with ur explaination.

  • @amaimask8685
    @amaimask8685 4 ปีที่แล้ว

    I have non cointegration data but when i change trend specification to be anything except beside constant i find cointegration, how to determine trend specification?

  • @dr35106
    @dr35106 5 ปีที่แล้ว +1

    You got a mazing teaching skill..thnx

  • @LearningEconomics
    @LearningEconomics 4 ปีที่แล้ว +1

    Thank you for your good explanations and helpful videos.
    I have a question. The objective of my research is to study the direction and magnitude of response of one variable due to a change in another variable. For that, I am planed to use SVAR and generating IRF. Pls tell me if variables are I(1) and there is cointegration then should I generate IRFs from VECM or from the unrestricted VAR? which approach is better?

    • @PatObi
      @PatObi  4 ปีที่แล้ว +1

      Here's my take but I'm not completely sure: If variables are I(1) but not cointegrated, the OLS regression we run is "unrestricted VAR
      ." In this case, we difference each series and then run OLS on the differenced series to examine their short-run dynamics. In your case though, the I(1) variables are cointegrated. So you can estimate (1) OLS regression on the levels data (unrestricted VAR) to examine the l.r. equilibrium relationship and (2) ECM, to examine the speed of adjustment to l.r. equilibrium

    • @PatObi
      @PatObi  4 ปีที่แล้ว +1

      The following ResearchGate Q&A might help: www.researchgate.net/post/Can_we_use_unrestricted_Var_model_if_the_variables_with_I1_series_are_not_cointegrated

    • @LearningEconomics
      @LearningEconomics 4 ปีที่แล้ว

      Thank you

  • @manalisharma9367
    @manalisharma9367 3 ปีที่แล้ว

    Sir I have 5 variables in my vecm and 4 cointegration equations . Can I proceed with the model and how to interpret each cointegrating equation

  • @rima9442
    @rima9442 3 ปีที่แล้ว

    Good day Dr,
    if the result for cointegration is not cointegrated for linear trend but cointegrated for Restricted linear trend, should we change the default setting of VECM from 3)linear trend to 4) Restricted linear trend when we want to run VECM?
    Thank you for your time if you read and comment on this DR.

  • @naveensood5982
    @naveensood5982 4 ปีที่แล้ว

    Can Impulse response function can be applied to VECM or it is applicable to VAR only??????

  • @MrHugosky1
    @MrHugosky1 4 ปีที่แล้ว +1

    Professor Pat Obi, I Iove your classes, and I have three questions for you. 1) In case one of the series has no unit root at level but the second one has it, and both of the series have no unit root at first difference. Are they candidates for a VECM model? 2) When you ran the VECM in the video, the R squared was 0.19. Is R squared or F statistic important when calculating the VECM? 3) I am currently running a VECM for two time series, but there is autocorreclation in the residuals. What would be the best way to tackle this issue? (10% percent of the data are outliers). Thank you very much, and if you teach virtual classes please let me know, so I can have one of your classes.

    • @PatObi
      @PatObi  4 ปีที่แล้ว +1

      If one variable is I(0) while the other is I(1), you should not run a VECM. Instead, run ARDL (I have a 6-part video series on ARDL here on TH-cam). Don't worry too much about R-sq and F at this stage. Instead, focus on whether the series are cointegrated. Also, examine the short run dynamics regardless of whether the two series are cointegrated. Residual correlation is definitely not good. Try using different lags. Also, you could try different data structure (daily, weekly, monthly, quarterly, annually). One might also consider adding another regressor. Hope this helps! Thanks for subscribing to my channel.

    • @MrHugosky1
      @MrHugosky1 4 ปีที่แล้ว +1

      Thank you very much, for all your answers, Professor@@PatObi! I am going to check the ARDL videos.

  • @Azam_Pakistan
    @Azam_Pakistan 7 ปีที่แล้ว

    Great but what happened to the C3 and C4 that came out insignificant and you started carrying out tests on them to point to the problem? Secondly do we report the long run coefficients from the long run equation the way we report in case of OLS ?

  • @yassineounnabi4396
    @yassineounnabi4396 6 ปีที่แล้ว

    I think that you must take the p-1 lag in the VECM estimation, which is 2(var lag) - 1 = 1. Thank you for your efforts.

    • @alphonsedismasorango8406
      @alphonsedismasorango8406 6 ปีที่แล้ว

      What happens if the optimal lag is p=1? In estimating VECM do we still subtract the lag so that we get p-1=1-1=0?

    • @yassineounnabi4396
      @yassineounnabi4396 6 ปีที่แล้ว

      @@alphonsedismasorango8406 yes absolutely. In this case, you will take zero as the VECM's lag. Good luck.

    • @alphonsedismasorango8406
      @alphonsedismasorango8406 6 ปีที่แล้ว

      @@yassineounnabi4396 , thank you for the quick response. However, when I do that, it gives me an 'illegal lag specification' error. What should I do? Thanks again for your response.

    • @yassineounnabi4396
      @yassineounnabi4396 6 ปีที่แล้ว

      @@alphonsedismasorango8406 in the lag criteria for the var, try to enter 4. Then I think that the criteria chosen will be 3. Then, try to estimate the VECM by 2 lags. If this strategy doesn't work, try to change the number of lags tested in the VAR. Try this and tell me the results.

    • @alphonsedismasorango8406
      @alphonsedismasorango8406 6 ปีที่แล้ว

      @@yassineounnabi4396 , thanks. do we use the AIC, SIC or others in selecting the optimal lag? My AIC indicates one lag same as the SIC. Kindly advice.

  • @denistiyo7193
    @denistiyo7193 4 ปีที่แล้ว

    The lecture has been most helpful. However, I wish to know how one would conduct a wald test when they used one(1) lag and there are two explanatory variables?

    • @PatObi
      @PatObi  4 ปีที่แล้ว

      Go to Coefficient Diagnostics, click on Wald Test and specify all the short-run coefficient numbers, e.g. C(3)=C(4)=0.

  • @reyazmalik.ph.d.7734
    @reyazmalik.ph.d.7734 6 ปีที่แล้ว +1

    thank you for wonderful teaching sir... it helps a lot..

  • @accountlastname1170
    @accountlastname1170 4 ปีที่แล้ว

    thanks a lot for this! question: what does it mean when C(4) and C(5) have high p-values (insignificant) in the ECM but indicate granger causality in the wald test?

    • @PatObi
      @PatObi  4 ปีที่แล้ว

      Sorry, not sure. I'm sure there's a good explanation for that anomaly.

    • @accountlastname1170
      @accountlastname1170 4 ปีที่แล้ว

      @@PatObi In this example, would we say that the VECM is poor given the very low adjusted r squared?

  • @maurogrovermamaniluna6829
    @maurogrovermamaniluna6829 2 ปีที่แล้ว

    a question if there are two variables, the result is also two equations of vecm because only the first equation is found and why not the second equation, is it not possible to make an inference?

  • @makaisapitjiumbirua9980
    @makaisapitjiumbirua9980 6 ปีที่แล้ว +1

    thank very much , but i have encountered this problem , all my C(1).........c(7) , are insignificant , what should i do now?
    Coefficient Std. Error t-Statistic Prob.
    C(1) -0.030998 0.024244 -1.278581 0.2218
    C(2) -0.082780 0.360631 -0.229543 0.8218
    C(3) -0.030502 0.341345 -0.089357 0.9301
    C(4) -0.107738 0.297568 -0.362061 0.7227
    C(5) 0.356619 0.301433 1.183077 0.2565
    C(6) -0.012327 0.236723 -0.052073 0.9592
    C(7) 0.200735 0.224905 0.892533 0.3872
    C(8) 3087.805 1536.751 2.009307 0.0642

    • @PatObi
      @PatObi  6 ปีที่แล้ว +1

      I believe those are the short run coefficients. Testing them jointly using Wald test will confirm that in the s.r., your X-variable does not Granger-cause your Y variable. That's what I think :-)

    • @ranaijaz6584
      @ranaijaz6584 4 ปีที่แล้ว

      @@PatObi please can you send me your mail and whatsapp number.
      thank you very much.

    • @ranaijaz6584
      @ranaijaz6584 4 ปีที่แล้ว

      @@PatObi my whatsapp number 00818602552355
      ijazyounis@njust.edu.cn

  • @geetanjali3436
    @geetanjali3436 3 ปีที่แล้ว

    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

    • @PatObi
      @PatObi  3 ปีที่แล้ว +1

      Some of the remedies may include (not guaranteed) increasing sample size, changing the lag structure, changing data frequency, e.g. from yearly to quarterly or monthly, etc.

  • @chhitijpoudel156
    @chhitijpoudel156 4 ปีที่แล้ว

    if non- stationary time series are integrated of the first order I(1) found not co-integrated , then what should we do ??

    • @PatObi
      @PatObi  4 ปีที่แล้ว

      Test for only short run causality.

    • @chhitijpoudel156
      @chhitijpoudel156 4 ปีที่แล้ว

      @@PatObi do you have video for that one ?

    • @PatObi
      @PatObi  4 ปีที่แล้ว

      @@chhitijpoudel156 Yes. Watch from about the 8 minute till end. It's part of the VECM. Also, check out my 6-part ARDL series. One of parts is on s.r. dynamics.

    • @chhitijpoudel156
      @chhitijpoudel156 4 ปีที่แล้ว +1

      @@PatObi thank you so much sir!! 🙏

  • @josephwu3172
    @josephwu3172 6 ปีที่แล้ว +1

    thx for making this video!!! it helps a lot

  • @anubhavagarwal236
    @anubhavagarwal236 3 ปีที่แล้ว

    Dear sir,
    In the video where you estimated an unrestricted VAR, you suggested that 2 lags were optimal to use based on SIC criteria. If I understand correctly, one must use (p-1) lags for Johansen's test and for VECM (where p is the optimal number of lags suggested by unrestricted VAR). But then you use 2 lags for both of these instead of 1. I am a little confused about that. Thank you in advance!

  • @chhitijpoudel156
    @chhitijpoudel156 4 ปีที่แล้ว +1

    Thank you so much !!!! i'm gonna graduate soon!!! :)

  • @전지홍-y9v
    @전지홍-y9v 5 ปีที่แล้ว +1

    Thank you for your good video.

  • @TehMinijimbo
    @TehMinijimbo 4 ปีที่แล้ว

    For the VECM, would you not have to drop from 2 lags to 1?

  • @gakidema3487
    @gakidema3487 5 ปีที่แล้ว

    Hello sir
    i have some questions on the error correction term. as u have mentioned in your lecturer that if the coefficient of the error term does not fall within -1 and 0 then there is problem with the model . in such case what is your suggestion to correct the model. I have my error term -2.00 and am confused about how to deal with it.
    i tried to reduce the lag length of my model from 4 to 3 and the error term falls within the range. if i do this step will it be a problem or is it ok to reduce the lag length. I have a sample of 30 observations and when i generate the optimal lag, the model gives me 4 lag.
    if you could clarify on this.
    thank you

  • @tyronedealwis5978
    @tyronedealwis5978 3 ปีที่แล้ว

    Dear Prof. I am Tyrone from Sri Lanka once again. How could I start learning time series model fully. I have no basic knowledge . would you advise me on this
    tyrone from Sri Lanka

    • @PatObi
      @PatObi  3 ปีที่แล้ว

      This 4-part series is a good start

  • @jackylin6281
    @jackylin6281 4 ปีที่แล้ว

    Hi professor. My three variables are all I(1). According to Johensen test, there are 2 cointegrations. However,when I construct VECM, C(1) is not significant and sometimes not negative.
    I have no idea where the mistake exists because I followed your method step by step.

    • @PatObi
      @PatObi  4 ปีที่แล้ว

      Please watch from the 12th minute for some ideas

  • @jayasubha96
    @jayasubha96 6 ปีที่แล้ว +1

    THANK YOU SO MUCH SIR FOR MAKING THE CONCEPTS SO CLEAR. I HAVE DID AN EXAMPLE IN WHICH MY RESULTS ARE QUITE GOOD EXCEPT THAT MY ESTIMATED MODEL PROBABILITIES WERE GREATER THAN 0.05.SO ANY OF THE COEFFICIENTS WERE NOT SIGNIFICANT. SO, WHAT IS THE INTERPRETATION? COULD U GIVE ME A SUGGESTION, SIR????

    • @PatObi
      @PatObi  6 ปีที่แล้ว

      jaya subha: If the short run coefficients of the explanatory variable are jointly not significant, then there is no short run causality.

    • @jayasubha96
      @jayasubha96 6 ปีที่แล้ว

      thank u :)

  • @xpattv
    @xpattv 4 ปีที่แล้ว

    Thanks Pat for this video. My question is what If i get cointegration in trace statistics but not in Max Eigenvalue statistics? can I proceed to the VECM model with such results?

    • @PatObi
      @PatObi  4 ปีที่แล้ว

      I would. Perhaps you might find evidence of l.r. causality. But be sure to ask others.

    • @xpattv
      @xpattv 4 ปีที่แล้ว +1

      @@PatObi Much obliged.

  • @w.omarfadhliw.mahmudkhairi6824
    @w.omarfadhliw.mahmudkhairi6824 5 ปีที่แล้ว

    If i am not mistaken, VECM must have -1 VAR lag numbers. If you choose VAR lag 2, so VECM wll be lag 1. Right?

  • @amritasengupta4260
    @amritasengupta4260 3 ปีที่แล้ว

    If the coefficient of the ECT is positive and significant, then can we say that there is long run causality? Because the coefficient is significant?

    • @PatObi
      @PatObi  3 ปีที่แล้ว

      No. Positive ECT coefficient means there's no convergence.

  • @iroegbuteuchendukelvin1925
    @iroegbuteuchendukelvin1925 5 ปีที่แล้ว

    Prof ,thanks for your explanations , from the VECM how do i determine my speed of adjustment at long-run. Thanks you.

    • @ImperiumLearning
      @ImperiumLearning 5 ปีที่แล้ว

      The speed of adjustment for each variable can be calculated by multiplying the coefficient of the variable in the cointegrating equation by the coefficient of the ECT where the first-difference of that variable is the dependent variable
      . If you don't understand what I mean: th-cam.com/video/FvCuHqqdasc/w-d-xo.html

  • @glenborg7884
    @glenborg7884 7 ปีที่แล้ว +3

    Thats amazing, thank you kind sir.

  • @sabirekose1601
    @sabirekose1601 5 ปีที่แล้ว +1

    this was really helpful, thanks a lot.

  • @dr.mohammadchhiddikurrahma4972
    @dr.mohammadchhiddikurrahma4972 7 ปีที่แล้ว

    Dear sir, can you please tell me how can I write asymmetric Vecm model after VECM analysis?

  • @Azam_Pakistan
    @Azam_Pakistan 6 ปีที่แล้ว

    Sir please add video on multiple structural breaks and how to deal with these.

  • @ousmanediallo535
    @ousmanediallo535 5 ปีที่แล้ว +1

    well explained, thanks man

  • @mostafaaboelsoud2734
    @mostafaaboelsoud2734 6 ปีที่แล้ว

    Dear Dr. Pat Obi. Thank you so much for such videos .. you have helped me a lot to finish my recent research paper. However, may I ask which software did you use to capture your lecture? I am really interested in recording my lectures as well, but I don't know the way!

  • @anupamsabharwal4385
    @anupamsabharwal4385 4 ปีที่แล้ว

    getting an error of insufficient number of observations while running VECm, kkndly guide sir

  • @rethabilemolapo1657
    @rethabilemolapo1657 3 ปีที่แล้ว

    What should you do when your results have no cointegrating relationship

    • @PatObi
      @PatObi  3 ปีที่แล้ว +1

      Then check only for short-run relations. No cointegration means no long run relationship.

    • @rethabilemolapo1657
      @rethabilemolapo1657 3 ปีที่แล้ว +1

      @@PatObi Thank you. This video really helped.

  • @abdullahiosman8654
    @abdullahiosman8654 ปีที่แล้ว

    why don't you reverse the sign of the coefficients?

  • @mishalkhaled8327
    @mishalkhaled8327 5 ปีที่แล้ว

    Sir thank you for your excellent explanations. I just confuse cos I got one variable (logTT) zero coefficient for long run how I explain it? can i remove it from the equation? can we use your model for more than 2 variables? thanks a lot
    Vector Error Correction Estimates
    Date: 02/02/19 Time: 20:53
    Sample (adjusted): 2007M04 2017M11
    Included observations: 128 after adjustments
    Standard errors in ( ) & t-statistics in [ ]
    Cointegrating Eq: CointEq1 CointEq2
    LKSEI(-1) 1.000000 0.000000
    LOG(TT(-1)) 0.000000 1.000000
    LHINDEX(-1) -0.154923 -1.038932
    (0.25311) (0.08584)
    [-0.61208] [-12.1034]
    LOG(INVP(-1)) 0.435698 0.245088
    (0.18477) (0.06266)
    [ 2.35810] [ 3.91135]
    C -11.12377 -4.368917

  • @gedeaolocks155
    @gedeaolocks155 7 ปีที่แล้ว +1

    thank you so much!!

  • @Dao-tao-Sale-Marketing
    @Dao-tao-Sale-Marketing 3 ปีที่แล้ว +1

    Thanks

  • @amaimask8685
    @amaimask8685 4 ปีที่แล้ว

    Obi why in delta y i is 1 and in delta x i is 0

  • @mrvideos1934
    @mrvideos1934 6 ปีที่แล้ว

    Sir, In the Long Run Model, how we explain it ? Increase in oil prince positively affecting the FX ? Or the sign reversed in Ect .. Please help

    • @ImperiumLearning
      @ImperiumLearning 5 ปีที่แล้ว

      The sign is reversed in ECT. To see why this is, set ECT to equal 0. Eg. 1.00FX_t-1 + 0.003Oil_t-1 -1.0308 = 0. Therefore, 1.00FX_t-1 = -0.003Oil_t-1 + 1.0308. Just in case you're wondering why we set ECT to equal 0, it's because ECT should have 0 mean.