[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: modeling constraint question
I'm not sure if I understand your question, but let me try a couple of
questions for you.
Do you want to constrain your function that estimates the parameters
(what Nick is calling upstream)? Or constrain your predictions
(downstream)? Constraining the predictions would seem to just be a
matter of -replace-ing the predicted funding07 with the 10% limits you
mention on funding06, perhaps using the lag operator if you want to
condition the limits on a previous year's observations.
replace funding07 = (L.funding06)*1.1 if funding07>(L.funding06)*1.1
replace funding07 = (L.funding06)*0.9 if funding07<(L.funding06)*0.9
I would guess, though, that constraining the estimation is a better
approach---correctly modeling the data generating process. To
constrain the estimation, see the online and Reference Manual help for
On Apr 26, 2005, at 6:02 PM, William P. Kittredge wrote:
Does anyone know a way to constrain the predict function? I'm working
with a simple funding model that the organization is comfortable with:
funding06 = bx1 + bx2+ bx3+e
I'd like to constrain the funding07 values for each grantee so that
they are +/- a given percent (e.g. if the budget goes down 10% each
grantee gets 10% less.
The reason for doing it this way, I'm told, is that the variables
(e.g. x1) change differently from year to year for each grantee and
that the 'need' they represent is also supposed to alter their funding
(need based) all else equal.
* For searches and help try: