EViews.com

We are forming symmetric CIs under the assumption of asymptotic normality, even in the case of bootstrapping, as the latter is for the coefficient covariance and not for the coefficient distribution.

Funny, I'm not seeing that behavior. Estimation from 1953 to 2018 gives estimates with 789 observations for ARMA-ML. Dynamic forecasting for 2019 gives values starting at -1.04 and ascending gradually to -0.55 As to the behavior that you reported... I'm not saying that this is what happened, but not...

It should work. Can you give us a bit more detail on precisely what you are doing, what you expect, and what you are seeing.

Thanks.

Two parts to the question. First, you can add a confidence ellipse as a layer to the scatterplot. From the interface, click on the graph, and select the desired confidence ellipse in the Fit Line combo. In command form group grp1 x y grp1.scat cellipse There are options for the computation of the el...

You can reparameterize using the @logistic or other such transformation. Simply replace the current parameter, say C(5), with @logit(C(5)). EViews will estimate the deep unrestricted parameter, while the model sees the restricted value.

Unfortunately, computing forecast standard errors for MS switching is extremely difficult and not currently supported. Support will almost certainly have to be using simulation methods, which are likely to be computationally intensive. I'll put this on the list of things to consider for future devel...

I'm don't understand the request.

Sorry to take so long. Missed seeing this one.

It will be fixed in the next patch.

Not built-in. The nonlinear nature of the estimators makes this difficult.

Have to think about this. Can you send me the workfile?

Looking at the variance specs only, it looks okay to me.

It's not a built-in feature. Note that the residuals that we compute aren't regime specific but rather an expected value using the regime probabilities. This accords with the notion that you never know what regime you are in. For the computation that you described, you'd have to use the coefficients...

Hard to say. The switching objective is very different from the simple minimum sums-of-squares that you get in the standard case.

What is regime specific in your model?

My 2 minute look at the De Jong and Mackinnon (1988) paper suggests that if the above is indeed what you want, you can get it by saving the one-step ahead and filtered state covariances and doing a recursion. This is based on a quick look so I may be wrong, but it looks as though it's possible from ...

The former.

Do you want cov(state(t|T), state(T|T))?

I haven't looked at this in a while, but off the top of my head, I can't think of an easy way to get that one from what we provide you. I'll have to take a look at the papers again.