Markov Switching model with logistic transition probabilities

[asbatronic2016]

Hi

I want to estimate a MS model with two states for volatility:
Vol=a*Vol(-1)+e, where a can be high or low depending on the state. See eq1 in the WF
Then I want the transition probabilities to depend on lagged market depth (Depth(-1)). P governs the transition probability during normal time (low a) and Q during stress periods (high a), I want both P and Q to follow a logistic distribution but with different parameters:
P(depth(-1))=[exp(theta0+theta1*depth(-1))/(1+exp(theta0+theta1*depth(-1))]
Q(depth(-1))=[exp(theta2+theta3*depth(-1))/(1+exp(theta2+theta3*depth(-1))]

How do I tell Eviews to use the logistic distribution for the transition probability?

MS_example.WF1

(27.66 KiB) Downloaded 253 times

Thanks in advance

[EViews Glenn]

Isn't that what you are estimating in EQ2?

[asbatronic2016]

Yes it is but as far as I understand, in eq2 i am unable to tell Eviews that I use a logistic model for transition probabilities on depth. How do I specify that?

[startz]

In the Help system look at Switching Regression/regime probabilities

[asbatronic2016]

Thanks, now I get it. Do you have an idea why I get NA in eq2 when I add a constant?

[startz]

I don't get an NA. Is our version of EViews fully uptodate?

[asbatronic2016]

It is Eviews 8, I get NA when I add a constant in the transition prob box to reflect my model.

ms_example (1).wf1

(27.93 KiB) Downloaded 271 times

[startz]

I don't know whether the algorithm changed between 8 and 9. Do be sure that you've applied all the updates to 8...just in case.

[asbatronic2016]

Cool, looks like updating to Eview9 solved the problem!
Cheers