too lenient

[startz]

The command

Code: Select all

ls x c(-1) ma(1:48)

runs a regression with one MA term. It probably should report an error.

Possibly an error should also be reported on the "c(-1)" term. Or at least whoever typed "c(-1)" should be advised to stop working and go out for a drink.

[EViews Glenn]

I believe that the c(-1) is giving you the lag of the constant (intercept) series.

[startz]

Maybe. But it turns out that c(-1) affects the smpl. Also, the estimated coefficient ends up in c(1).

[EViews Glenn]

Yes it does, that's the point of me pointing out that it was the lag of the series. Note that what you are seeing is the effect of the symbol "C" having multiple interpretations. In a series construct, it's simply the "ones" series. Note that you can add "C" to a group and it will give you a series containing the value 1 for all observations. And as you note, in an object context, it's the coefficient vector.

I know that it's a bit odd, but it's a holdover from the early TSP days and at this point, it's not going to change.

[startz]

Okay, I understand now. It's David's fault.

[EViews Glenn]

I don't think fault is the correct characterization.

[startz]

I accept the correction. "Feature?"

[EViews Glenn]

"Behavior"