How to retrieve regime specific AR coefficients while estimating a univariate Markov regime switching model?
please refers to Table I. Univariate regime switching models for stock and bond returns of An Econometric Model of Nonlinear Dynamics in the Joint Distribution of Stock and BondReturns by Massimo Guidolin and Allan Timmermann (2006) for additional info.
How retrieve AR coeffecients in Markov model (MSIAH)
Moderators: EViews Gareth, EViews Moderator
-
- Posts: 2
- Joined: Thu Jun 29, 2017 7:24 am
-
- EViews Developer
- Posts: 2671
- Joined: Wed Oct 15, 2008 9:17 am
Re: How retrieve AR coeffecients in Markov model (MSIAH)
Have you estimated the model already?
-
- Posts: 2
- Joined: Thu Jun 29, 2017 7:24 am
Re: How retrieve AR coeffecients in Markov model (MSIAH)
Hello Glenn,
Yes, I have. I am a noob and thus I'm not sure the extent to which my estimation is accurate.
I am following a study that is very much similar to https://doi.org/10.1016/j.pacfin.2013.12.003 and is required to conduct a univariate and multivariate Markov regime switching estimations. I have attached a screenshot of parameters I've set while estimating and haven't tinkered with the option tab. The Independent variables used are the same as of the journal mentioned above. The dependent variable is percentile return on stock market of 6 GCC countries. {In picture, KSA (Ticker: TASI).}
Also, how do I conduct the Regime classification measure (RCM) Test and Davies test on Eviews? I could not find any link for help.
Furthermore, I sometimes get ' NA ' for 'for Std. Error, z-Statistic and Prob. on estimation with 3,4-State MSIAH with AR(1) for some countries. I also got 'NA on a 2-State MSIAH with AR(1) [it only had 107 estimations tho].
Thank you!
Yes, I have. I am a noob and thus I'm not sure the extent to which my estimation is accurate.
I am following a study that is very much similar to https://doi.org/10.1016/j.pacfin.2013.12.003 and is required to conduct a univariate and multivariate Markov regime switching estimations. I have attached a screenshot of parameters I've set while estimating and haven't tinkered with the option tab. The Independent variables used are the same as of the journal mentioned above. The dependent variable is percentile return on stock market of 6 GCC countries. {In picture, KSA (Ticker: TASI).}
Also, how do I conduct the Regime classification measure (RCM) Test and Davies test on Eviews? I could not find any link for help.
Furthermore, I sometimes get ' NA ' for 'for Std. Error, z-Statistic and Prob. on estimation with 3,4-State MSIAH with AR(1) for some countries. I also got 'NA on a 2-State MSIAH with AR(1) [it only had 107 estimations tho].
Thank you!
- Attachments
-
- Parameters set for Univariate Non-AR estimation
- Univariate Non AR .png (88.21 KiB) Viewed 4141 times
-
- Parameters set for Univariate estimation
- Univariate .png (93.64 KiB) Viewed 4141 times
-
- Parameters set for Multivariate estimation
- MultiVariate.png (103.15 KiB) Viewed 4141 times
-
- EViews Developer
- Posts: 2671
- Joined: Wed Oct 15, 2008 9:17 am
Re: How retrieve AR coeffecients in Markov model (MSIAH)
Still not sure what the question is...
The coefficients of the model are available in the usual way as an @coefs data member off of the estimated equation.
The only classification measures we report at the regime probabilities, which you can use in various ways to compute various RCMs. The tests to which you refer are not built-in, nor is the Cho-White test.
The singularities you report are common with MS models. One often has to play around with starting values, but even so, some models are not be easily estimable.
The coefficients of the model are available in the usual way as an @coefs data member off of the estimated equation.
The only classification measures we report at the regime probabilities, which you can use in various ways to compute various RCMs. The tests to which you refer are not built-in, nor is the Cho-White test.
The singularities you report are common with MS models. One often has to play around with starting values, but even so, some models are not be easily estimable.
Who is online
Users browsing this forum: Google [Bot] and 28 guests