Switch Regression Model Problem

[EViews Gareth]

Well I guess you have to ask yourself why you disabled automatic updates :D

At this point, given the mess you're in, you're probably best just uninstalling and re-installing with the latest installer.

http://www.eviews.com/download/ev8downlinks.html

[nicotrader]

I finally solved the problem.
Unfortunately it has presented a new problem:

pretty much when i go to select the sample es. 1 5000, just end and the output is displayed is known that the model was estimated over the whole sample and not just on the range
1 5000! Why??? In switch regression it is not possible to select a different sample from whole series loaded?

[EViews Gareth]

Switching regression allows samples in the same way as every other type of estimation in EViews.

[nicotrader]

Ok so why not let me select a sample?

[nicotrader]

the full data set is of 54993 observations. I selected a sample 1 5000 but the output returns the entire range. Why!?

[startz]

Perhaps you could post the exact commands, showing how you selected the sample. Also, check the sample specification in the equation estimation dialog.

[EViews Gareth]

There is nothing in that screenshot that indicates you used a sample of "1 5000"

[nicotrader]

[EViews Gareth]

Could you provide the workfile?

[nicotrader]

Now it works is probably a bug. Sometimes it works sometimes not, changing the series.

But I have another question:

if i value on a sample of 5000 when i go to estimate the model the odds are up to 5000. At this point if i do forecasting, however, it does not allow me to extend the odds in any standard but only the result of regression. How can I do?

[nicotrader]

Any advice??

[EViews Gareth]

Could you rephrase your question so that it is understandable?

[nicotrader]

Why the probability of an estimate Markov switches are not testable out of sample but only in sample?
What good has test only in sample?

[startz]

Why the probability of an estimate Markov switches are not testable out of sample but only in sample?
What good has test only in sample?

It isn't clear exactly what you want to do. What do you want to compare out-of-sample estimated Markov switches to?

[nicotrader]

I would like to test in sample and out of sample the state probabilities