TAR-MTAR - Singular Matrix

[eficefic]

Im trying to run the following regression but a "near singular matrix" message keeps popping up. I'ld appreciate any suggestion or input to solve this issue:

y2 = y2(-1) + error_whitenoise1
y1 = y2(-1)
I_tar = (y1 >= 0)

freeze(yuroot) eq01.ls d(y2) I_tar*y1 (1-I_tar)*y1 c


*This is meant to create a data generating process with the following structure:

d(y)=rho1*y(-1)*dummy+rho2*y(-1)*(1-dummy)+error, where dummy is "I_tar" for the TAR model.

Thanks in advance!

[EViews Gareth]

Does I_tar have variance?

[eficefic]

I_tar is a dummy = 1 when y(-1)>=0, and 0 when y(-1)<0

[EViews Gareth]

Sure, but does it actually have variance? I mean is it the case that, in your estimation sample, it only has 0s or only has 1s?

[eficefic]

Nope. It is always 1

[EViews Gareth]

Then (1-I_tar)*y1 is perfectly correlated with the constant.