Problem with lags in SETAR model

[statistik123]

hey everybody

i have a problem with the estimation in setar model in eviews 9. i want to estimate a univariate time serie with only one dependend variable. but every time when i enter a lag to the equation there is a error message.
can anybody help me?

[EViews Gareth]

How are you specifying the lag, and what is the error?

It is hard to say what you're doing wrong if we don't know what you're doing

[statistik123]

i have one time series and to see which delay parameter, threshold parameter, regimes und lags i take, i want to try different combinations and check which one minimizes the AIC.
So i have the dependend variable from my time serie and first want to create 1-4 lags with -1 to -4.

error messages like:

Near singular matrix error. Regressors may be perfectly collinear.

or

1TO is not defined.

but i don´t know, what i have to change cause i´m new with eviews ;)

[EViews Gareth]

Just enter "1 4" without the quotes.

[statistik123]

It´s the same. When I wrote the dependent variable without space: (dependent variable1 4) then there is the error: REGIME_11 is not defined.

When i wrote it with space (dependent variable 1 4) then the error is: Near singular matrix

[EViews Gareth]

Just enter "1 4" without the quotes.

[statistik123]

I entered it without quotes like:

dependentvariable 1 4

but it doesn´t work...

[EViews Gareth]

Don't enter dependentvariable.

[statistik123]

Number of unique values of the threshold variable 1 are less
than the maximum number of thresholds requested.

I don´t get it...

[EViews Gareth]

I think you have two options, you can either read through the documentation properly:
http://www.eviews.com/help/helpintro.ht ... .3.html%23
https://www.youtube.com/watch?v=tLuE0oMAQ_o

Or you can give precise, full details on exactly what you have done so that we are able to tell what you've done wrong.

[statistik123]

Okay, generally I have to create a forecast from a nonlinear, univariate time series.
Therefore I have only one time serie from an financial indicator which is my dependent variable.

I start with the estiation of a SETAR model and have to test which variables (delay, threshold, lags) create the best model (which minimizes AIC).
At the beginning I try for example the delay from 1 to 4 and lags from 1 to 4, so I have 16 different combinations and I also split my variable in two regimes (Regime1 and Regime2).

But no matter what lag combinations I enter, there is a error. I think there is something wrong in the Equation specification.

I tried different combintions, but what is wrong when I try it like:

Dependent variable followed by list of threshold regressors:
indicator 1 4
List of non-threshold regressors:
c
Threshold variable specification
2

[EViews Rebecca]

What are your threshold regressors (what are the variable names)?

[statistik123]

it´s the same like the delay and lags. in every estimation i have to assume a threshold variable and check which one minimizes the AIC (thats the procedure which my prof preferes).

[EViews Rebecca]

What are the variable names of your threshold regressors? Are you trying to test all delays from 1 to 4, or one at a time? You indicate both in your previous posts.

[statistik123]

I have no names of threshold regressors, I only have the name of my variable (indicator) and regime1 and regime2 as the split of my dependent variable. I try to test all delays from 1 to 4, but each in every estimation. So in my first estimation, the delay is 1. In the second the delay is 2 and so on.

Sorry for the confusions :)