Thank you Justin for the recap I was working on my project it was very to recall thank you for the service
@MrRobertLake
3 жыл бұрын
Some serious issues here. 1) The Johansen's method build on a fIML stimation of eigenvalues. FIML is extremely sensitive to misspecification. Therefore one has to test each equation in the var for NID residuals. Furthmore you MUST investigate the series for outliers and put in impulse dummy variables to achive NID. Under no circumstances can you let the program decide on the lag length only on information citieria. Never trust the results of a VAR that has not been tested properly. 2) Of course all variables should be I(1). This needs to be tested in advance. Otherwise you should test in Johansen's VAR. If one variable is I(0) you automatically get one significant eigenvalue and cointegrating vector. 3) You cannot include an I(2) variable either since that will also ruin the test. In you model you include CPI that can be an I(2) variable. 4) Johahansen's test uses the trace statistics. The other statistic has proved inconsistent. Not many people know that. 5) The output of the eigenvalues from the cointegration test comes in the wrong order. The testing need to follow the Pantula principle. So you start with zero significant eigenvales. If that is rejected you test for at least one sign eigenvalue etc. When you cannot reject you must stop. After that the results are no longer valid. The fact that you find 2 sign eigenvalues among these three variables is likely to depend on lack of testing the individial series and te VAR model. Please don't use R for testing cointegration. You lose so much time on basic syntax. For this work gretl is better, and of course PcGive with CATS is the best. I recommed "A Guide to testing for Unit roots and cointegration" on researchgate.net.
@ideas.with.justiin
Ай бұрын
Thanks Justin
@周育呈-l6w
Жыл бұрын
you saves my life, thank you!!!
@adilmasoodkhan4527
3 жыл бұрын
Sir, we will subtract one from the lags selected by AIC or any other criterion while we will be estimating VECM. But why have you subtracted one in the Johansen cointegration test itself?
@FR17509
4 жыл бұрын
Great work! Thank you very much! How should we interpret the resulrs? We found two cointegrating relationships, but how do they look like and how should we interpret them? Is it important to know this? I guess it means, we found a long-run relationship between the variables.
@JustinEloriaga
4 жыл бұрын
We will use the VECM to explain that more.
@pakwidi531
9 ай бұрын
wah... thank you so much 😁🙏
@getinenglish3472
3 жыл бұрын
Bro, well done to you. This is such a great video! just one question. What implications would I have if the Eigen test does not accept the null hyp in any case but the Jo trace does? IS it still ok to carry on with the model? Thank you :) Andre
@palashsrivastava7711
4 жыл бұрын
Hi Justin. Thank you very much for the video. My question is do we use the Cointegration Test instead of Granger Causality if we are doing VECM? Also when can we expect video on VECM? Thanks again!
@JustinEloriaga
4 жыл бұрын
Hello! Thanks for your comment. Video on VECM will most likely be up at the end of the week. You may also opt to do a Granger causality test in addition to a Cointegration test in VECM. Hope this helps.
@palashsrivastava7711
4 жыл бұрын
@@JustinEloriaga Thank You. Looking forward for that video.
@johnarvinbernabe1767
4 жыл бұрын
Thank you very much for this, Justin. May I know if you have a video uploaded on VECM?
@sampejuan0173
3 ай бұрын
how do i interpret the cointegration relations below?
@valentinadidenko6253
3 жыл бұрын
Thank you very much! What is my mistake? I haven't Johansen summary ( summary(ctes1t) Length Class Mode 1 ca.jo S4
@fly77x
3 жыл бұрын
Thank you. Then, the johansen test uses a VAR model implicitly?
@omarkanaan7515
3 жыл бұрын
Awesome! Thank you :D
@adilmasoodkhan4527
3 жыл бұрын
Sir, since we get two cointegrating relationships. So here there will be having two Error correction terms. Then how do we interpret these two? Or we need to make some conversion and make it compatible to interpret. Since I don't know, how will we be interpreting two ECT per quarter? plz, help me to understand.
@Fightclub1995
3 жыл бұрын
I keep getting the error message insufficient degrees of freedom. Do you know what this could mean?
@raulq.3519
4 жыл бұрын
Justin, is it possible to run a Johansen test using 1st differences? Or only variables in level/ln?
@JustinEloriaga
4 жыл бұрын
Yes, I believe it should be possible. Hope this helps.
@adilmasoodkhan4527
3 жыл бұрын
@@JustinEloriaga But Sir, the cointegration should be run at level series because if we difference the series then we lose the long-run information, and probably we might not be getting cointegration
@thiagobrito208
3 жыл бұрын
Suppose you have two time series. We can say that they're cointegrated if there is a linear combination between these series that results in a I(0) ts. If you run this cointegration test using 1st differenced series, then you'll be using two stationary processes. In this case, any linear combination of these processes will result in another stationary process. In other words: Johansen test will return K cointegration relationships (where K equals to the number of series involved) every time. As far as I know, there is no point in running this.
@DanielaSantos-jj8vg
3 жыл бұрын
I’m having problems in doing that, can you help me please?
@OutperformMP
3 жыл бұрын
If VARselect suggest 1, it means that I should choose 0 as lag. But the ca.jo test requires that K is at least 2. What does this mean? What should I do then?
@robertkimcabeza3038
3 жыл бұрын
Good evening sir, I would like to ask Po, saan Po galing si "lagmax=7" sa line 25 po sa program. ?. Sana Po ma notice tong tanong ko. Thank you
@afiqaiman5874
2 жыл бұрын
is it possible to use ca.jo function if my specified lag length 1?
@catsworld2828
2 жыл бұрын
Should the Johansen Cointegration test by conducted on variables at levels or in their transformed state?
@EvsEntps
2 жыл бұрын
At levels, because the JC test establishes whether a linear combination of 2 or more I(1) time series produce a I(0) (i.e. stationary) series. If you do it on I(0) series the test is redundant because you already have I(0) series and you will always be able to find a linear combination that will be I(0). If you do get cointegration between 2 or more I(1) time series then you implement a VECM version of the VAR model where instead of using those series you use their cointegration relations (the relevant R functions should handle this for you).
@aguilar3271
2 жыл бұрын
Thank you
@rafaeldemoraislima9080
3 жыл бұрын
Thanks sir! Is it ok if my R= 2?
@fvc1612
2 жыл бұрын
Mmm why do you only use 7 lags?
@bobbybaylonjr7403
4 жыл бұрын
Hi, thank you so much for this. Hope you could share the data and the code. I would like to follow along. The link you provided is not working. many thanks and great job!
@JustinEloriaga
4 жыл бұрын
Hi, thanks for your comment! Here is the link: drive.google.com/drive/folders/1QHUwS83GAUOVnxnoi-OZ6-shmTwo8mdB?usp=sharing
@palashsrivastava7711
4 жыл бұрын
Any specific reason to use log values? Is it mandatory?
@balautkarsh
3 жыл бұрын
log operation streamline the values. eg cpi and gdp are in 2 digits and 10 digits, applying log reduces them to single digits values, hence easily comparable. also, log is a monotonic transformation, hence doesnt affect economic properties.
@DanielaSantos-jj8vg
3 жыл бұрын
I’m having problems in doing that, can you help me?
Пікірлер: 37