Hey, if you liked the video, please subscribe, share and thumbs up. It would be great to be able to monetise my videos to the point where I could purchase some proper screen recording software. Also, if you have any questions about the video or if anything is unclear, don't hesitate to comment and ask me questions - I'll try to reply as soon as possible. Thanks.
Hello Imperium Learning, Firstly thank you very much for providing these series, of great help for under-graduate students (will make sure to share will other students). Secondly, Are we not suppose to select the (p,q) parameters of a GARCH model (or other related GARCH models (EGARCH...)) using Information criterion (AIC BIC)? Meaning that the selection of different orders in the parameters (p,q) are assessed on relative minimisation of AIC BIC criterion. Thank You for your answer
Hi Paul, thanks a lot for the kind feedback. That's a really good question. I think the key point to make is that a GARCH (1,1) model is the most parsimonious. By parsimonious, I mean the conditional variance equation contains the fewest possible terms. If your model satisfies all the residual tests that you apply (e.g. Ljung-Box and ARCH LM test), then that's the best possible model since it's well-specified (you'll know your model is well-specified when it passes all residual tests) and contains the fewest possible terms (most parsimonious). I think you may be able to minimise an information criterion to find the optimal number of parameters, but I would only do this if a GARCH (1,1) model doesn't satisfy residual tests. Bear in mind though, if minimising an information criterion leads to a high number of parameters, that might not be the optimal model. Hope that helps
Hey there, Thanks for the positive feedback. You're not the first person to ask for the PPTs but unfortunately I deleted all of them a year ago. I've since learned from that mistake since for the Excel course I'm currently in the process of uploading I'm also using Google Drive to upload all relevant documents.
Thank you for useful guiding video. And I have a question: my data set has ARCH effect but then, the alpha of ARCH term in GARCH(1,1) is negative, what should I do?
Good evening sir thanks a lot for such great video But I need more help in this regarding control variables. How we should apply garch 1,1 model if are also taking some control variables to control their effect on our variable. Please upload a short video dealing with control variables. It will be a great help for me.
Hey there, Don't forget to subscribe, it always helps a small channel like mine. Also, don't forget to share/recommend my channel to anyone else you might know. Addressing your question, I go into more depth on those two residual tests in my ARMA model videos (there are 3 of them). Hope that helps.
Great material again, thank you very much. Could you assist in detailing what the hypothesis test for each of the test you conduct? i.e., what would our GARCH hypothesis test be so forth for GJR, EGARCH...
look at each of the coefficients for each conditional variance equation. The GARCH term has coefficient denoted by beta at the beginning. Null would be that beta = 0, alternative would be that beta does not = 0 (and is, therefore, significant).
@@lawjef There was no serial correlation in my initial ARIMA model. Serial correlation emerged after adding the ARCH/GARCH terms. Well I was looking at the correlogram q statistics, not the squared ones. I thought the squared ones are for any remainder ARCH effects. As you said maybe some degree of autocorrelation in GARCH model doesn't affect too much the results. Thank you very much indeed for taking the time to answer my question. I really appreciate it.
Hey there, Honestly, I don't think so. The reason is that the more time periods ahead that you forecast, the more errors (and accumulated errors) will be included in your prediction. I'd dsay the absolute maximum is 10 days but even then you're going to have significant error.
Hey there, Yes I'm aware the videos in my first playlist were pretty quiet. I've resolved that in the second playlist. This is also not going to be an issue in my upcoming MS Excel course which I should be uploading the next few weeks.
@@ImperiumLearning At the end of that video, you did mention of two approach to test for asymetry in GARCH model. I couldn't get the two method of testing that you mentioned clearly neither was i able to find that perticular video from list. Could you please mention those methods again??
@@AlagieBSowe Hi there, here is the relevant playlist: th-cam.com/play/PLUSOCDNlb8srWoY1E7d1LvJfzTRboRVyF.html There are two models you can estimate - EGARCH and GJR-GARCH. These are videos 13 and 14 in the playlist. However, it's good to conduct the Engle and Ng test as a way to justify estimating these two models in the first place (videos 11 & 12 in the playlist).
![[Live] : ONE FIGHT NIGHT 25 วันนี้!! "อเล็กซิส vs รีเกียน"](http://i.ytimg.com/vi/suVr9a6BM2g/mqdefault.jpg)
Hey, if you liked the video, please subscribe, share and thumbs up. It would be great to be able to monetise my videos to the point where I could purchase some proper screen recording software.
Also, if you have any questions about the video or if anything is unclear, don't hesitate to comment and ask me questions - I'll try to reply as soon as possible.
Thanks.
Really enjoyed it. Simple and clear instructions. Well done.
Hello Imperium Learning,
Firstly thank you very much for providing these series, of great help for under-graduate students (will make sure to share will other students).
Secondly,
Are we not suppose to select the (p,q) parameters of a GARCH model (or other related GARCH models (EGARCH...)) using Information criterion (AIC BIC)?
Meaning that the selection of different orders in the parameters (p,q) are assessed on relative minimisation of AIC BIC criterion.
Thank You for your answer
Hi Paul, thanks a lot for the kind feedback.
That's a really good question. I think the key point to make is that a GARCH (1,1) model is the most parsimonious. By parsimonious, I mean the conditional variance equation contains the fewest possible terms. If your model satisfies all the residual tests that you apply (e.g. Ljung-Box and ARCH LM test), then that's the best possible model since it's well-specified (you'll know your model is well-specified when it passes all residual tests) and contains the fewest possible terms (most parsimonious).
I think you may be able to minimise an information criterion to find the optimal number of parameters, but I would only do this if a GARCH (1,1) model doesn't satisfy residual tests. Bear in mind though, if minimising an information criterion leads to a high number of parameters, that might not be the optimal model.
Hope that helps
Hey thanks for the videos. Its clear and easy to understand. Can you please post me the link, where will I get the PPTs of all these?
Hey there,
Thanks for the positive feedback. You're not the first person to ask for the PPTs but unfortunately I deleted all of them a year ago.
I've since learned from that mistake since for the Excel course I'm currently in the process of uploading I'm also using Google Drive to upload all relevant documents.
Great video. I was wondering how to get all garch data before plotting the graph.
Thank you for useful guiding video. And I have a question: my data set has ARCH effect but then, the alpha of ARCH term in GARCH(1,1) is negative, what should I do?
Good evening sir thanks a lot for such great video
But I need more help in this regarding control variables. How we should apply garch 1,1 model if are also taking some control variables to control their effect on our variable. Please upload a short video dealing with control variables. It will be a great help for me.
Thank you, you just saved me from my final exam, I hope you can talk more about Lm test and box-Ljung in another video
Hey there,
Don't forget to subscribe, it always helps a small channel like mine. Also, don't forget to share/recommend my channel to anyone else you might know.
Addressing your question, I go into more depth on those two residual tests in my ARMA model videos (there are 3 of them).
Hope that helps.
Great material again, thank you very much. Could you assist in detailing what the hypothesis test for each of the test you conduct? i.e., what would our GARCH hypothesis test be so forth for GJR, EGARCH...
look at each of the coefficients for each conditional variance equation. The GARCH term has coefficient denoted by beta at the beginning. Null would be that beta = 0, alternative would be that beta does not = 0 (and is, therefore, significant).
Hi there! Really useful video as all of yours! What if I find autocorrelation in the residuals though? Is there any way to deal with it?
@@lawjef There was no serial correlation in my initial ARIMA model. Serial correlation emerged after adding the ARCH/GARCH terms. Well I was looking at the correlogram q statistics, not the squared ones. I thought the squared ones are for any remainder ARCH effects. As you said maybe some degree of autocorrelation in GARCH model doesn't affect too much the results. Thank you very much indeed for taking the time to answer my question. I really appreciate it.
Probabilities are significantly above 0.05. .....it does not suffer from what? I cannot hear that. Please help me to type words
It is possible to forecast in eviews the Garch model¿? for a least 100 days?
Hey there,
Honestly, I don't think so. The reason is that the more time periods ahead that you forecast, the more errors (and accumulated errors) will be included in your prediction. I'd dsay the absolute maximum is 10 days but even then you're going to have significant error.
CR7 prime
Thank you for the Video. It will also be very helpful if you could speak a bit louder please!!
Hey there,
Yes I'm aware the videos in my first playlist were pretty quiet. I've resolved that in the second playlist.
This is also not going to be an issue in my upcoming MS Excel course which I should be uploading the next few weeks.
@@ImperiumLearning
At the end of that video, you did mention of two approach to test for asymetry in GARCH model. I couldn't get the two method of testing that you mentioned clearly neither was i able to find that perticular video from list. Could you please mention those methods again??
@@AlagieBSowe Hi there, here is the relevant playlist: th-cam.com/play/PLUSOCDNlb8srWoY1E7d1LvJfzTRboRVyF.html
There are two models you can estimate - EGARCH and GJR-GARCH. These are videos 13 and 14 in the playlist.
However, it's good to conduct the Engle and Ng test as a way to justify estimating these two models in the first place (videos 11 & 12 in the playlist).