## Switching regression Problem

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12038
Joined: Tue Sep 16, 2008 5:38 pm

### Re: Switching regression Problem

How is what possible?

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

I try to explain myself better and more simply:

first estimation
dataset excel:

1 2 3 4 5 6 7 8 9

insample:

1 2 3 4 5

forecast:

1 2 3 4 5 6 7 8 9

second estimation same model
dataset excel:

1 2 3 4 5 6 7 8 9 10 11

insample:

1 2 3 4 5

forecast:

1 2 3 4 5 6 7 8 9 10 11

Then, DIFFERENT OUTPUT INSAMPLE

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12038
Joined: Tue Sep 16, 2008 5:38 pm

### Re: Switching regression Problem

EViews Gareth wrote:
I'll also point out that re-estimating the same equation on the same data will also, probably, give different estimates and then produce different forecasts, since a) you're using random starting values, and b) your estimates didn't converge.

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

In my work add a new observation in excel dataset day by day and every day refresh the excel dataset and estimate every time the model, but with same parameters and same in sample dataset.

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

a) you're using random starting values

Why gareth?

I use a precise in sample dataset, no random starting values

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

The in sample dataset its always the same.
the only thing that changes is the total dataset, in particular out of sample dataset which present new observation every time.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12038
Joined: Tue Sep 16, 2008 5:38 pm

### Re: Switching regression Problem

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

I still do not understand why you talk about random starting values

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

I make another clarification, if it is not possible to obtain a stable in-sample estimate, then the forecast function is totally useless.
you think?

startz
Non-normality and collinearity are NOT problems!
Posts: 3424
Joined: Wed Sep 17, 2008 2:25 pm

### Re: Switching regression Problem

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

starts the value are every time the same.
I never change the value of the series in sample.
Only add a observation out of sample.
The value of in sample are same always

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

Please look, the value are same always.

first estimation
dataset excel:

1 2 3 4 5 6 7 8 9

insample:

1 2 3 4 5

forecast:

1 2 3 4 5 6 7 8 9

second estimation same model
dataset excel:

1 2 3 4 5 6 7 8 9 10 11

insample:

1 2 3 4 5

forecast:

1 2 3 4 5 6 7 8 9 10 11

Posts: 63
Joined: Sat Jun 06, 2015 11:48 am

### Re: Switching regression Problem

I use a different model in eviews, ex garch but i have a problem only with switch regression.
In garch model and forecast the output is always same if i use the same in sample dataset.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12038
Joined: Tue Sep 16, 2008 5:38 pm

### Re: Switching regression Problem

nicotrader wrote:I make another clarification, if it is not possible to obtain a stable in-sample estimate, then the forecast function is totally useless.
you think?

Probably more accurate to say that it is not a good idea for you to use Markov Switching