Hi all,
So I have created a MS-AR(1) model in EViews 9.5 and I'm just trying to understand how some of the output is calculated.
This is really dumb and probably a simple question to answer, but I can't seem to get how the fitted values are calculated. I have tried calculating them myself and can't seem to get my calculations to match up. My process looks like this:
fity1 = c(1)+beta(1)*(y(-1) - c(1))
fity2 = c(2)+beta(1)*(y(-1) - c(2))
Then the fitted value is:
fity=P(S(t)= 1)*fity1+P(S(t)= 2)*fity2
NOTE: I'm using smoothed probabilities in my calculations (which is probably part of the problem).
Please help, my calculations are really close, but don't match up with those produced in the output.
THANK YOU!!!
Calculating fits of Markov Switching Model
Moderators: EViews Gareth, EViews Moderator
-
- EViews Developer
- Posts: 2672
- Joined: Wed Oct 15, 2008 9:17 am
Re: Calculating fits of Markov Switching Model
Why are you using the smoothed? One other issue is that if it's an autoregressive model, the weighting is much more complex as it requires the entire state probability, and the states are defined for multiple periods.
Re: Calculating fits of Markov Switching Model
Hi Glenn,
Thank you for the response.
I can't say why I chose the smoothed. I'm really just not sure how to calculate the fits. I understand the Smoothed probabilities take into account the entire information set while the filtered take into account info up to the given point in time, but I don't know how to incorporate these into (or if they are incorporated into) the calculation of the fits. I feel like the transition probabilities play a role in the calculation as well, but I can't sort it out and from the literature i've read it's difficult for me to decipher how to use these to do this calculation.
Please let me know if you can give me some guidance here.
Thanks,
- Keith
Thank you for the response.
I can't say why I chose the smoothed. I'm really just not sure how to calculate the fits. I understand the Smoothed probabilities take into account the entire information set while the filtered take into account info up to the given point in time, but I don't know how to incorporate these into (or if they are incorporated into) the calculation of the fits. I feel like the transition probabilities play a role in the calculation as well, but I can't sort it out and from the literature i've read it's difficult for me to decipher how to use these to do this calculation.
Please let me know if you can give me some guidance here.
Thanks,
- Keith
-
- EViews Developer
- Posts: 2672
- Joined: Wed Oct 15, 2008 9:17 am
Re: Calculating fits of Markov Switching Model
You want the filtered.
Note that with AR(1) you'll need to keep track of the two dimensional state vector. The one dimensional state variable probs will not be sufficient so you would have to compute the recursion yourself as we report only the latter.. Our documentation describes this computation in some depth and offers references.
Note that with AR(1) you'll need to keep track of the two dimensional state vector. The one dimensional state variable probs will not be sufficient so you would have to compute the recursion yourself as we report only the latter.. Our documentation describes this computation in some depth and offers references.
Who is online
Users browsing this forum: No registered users and 33 guests