better fitted values

[startz]

Consider the regression
ls e^2 c x
If you ask for the fitted values of e^2, you may get the message can't raise negative number to a non-negative power. Now it's true that if you try to forecast e you may run into difficulty, but e^2 is always straightforward. It would be nice if the forecast of the actual LHS variable could be special cased so that it always works.

[startz]

Consider the regression
ls e^2 c x
If you ask for the fitted values of e^2, you may get the message can't raise negative number to a non-negative power. Now it's true that if you try to forecast e you may run into difficulty, but e^2 is always straightforward. It would be nice if the forecast of the actual LHS variable could be special cased so that it always works.

(1) bump

(2) The forecast dialog even offers to forecast the actual LHS variable, and then fails.

[EViews Glenn]

I've seen this one before and am sympathetic, but to do the special casing you're asking for we'd have to write a separate forecast engine for just this case (static, dependent variable fit--note that you can't do it in any dynamic settings). This one is on our list, but there's a fair amount of complexity in there and since the user can do this by re-estimating after generating the dependent variable, this hasn't worked its way to the top of the list.

But I am sympathetic, really...