Sorry, I was thinking along different lines in terms of modeling. I
thought that you were looking to reformulate your economic model to
account for the correlation structure of the predictor weather variables.
As far as modeling the weather variables amongst themselves... you haven't
defined them for us, so I'm going to trust that you have constructed them
in such a way that they are sufficiently close to normally distributed to
use comfortably in the models and simulations. If your simulation is
really meant to be a time series (which probably makes sense in this
context), then you want to generate your weather variables with not just
the proper inter-correlation structure (which is all we discussed so far),
but with the proper autocorrelation and seasonalities. Someone else here
may have a better idea of how to do this than I do, but I think it sounds
like a vector autoregression, so you might want to look into PROC VARMAX.
-- TMK --
"The Macro Klutz"
On Thu, 10 Nov 2005 09:12:56 -0800, Pete Larsen <phlarsen@YAHOO.COM> wrote:
>Thanks for the quick response.
>I'm not sure I totally follow you (yet). Are you
>saying that the two steps I need to follow involve (in
>1) modeling precip against temp using proc mixed
>(precip is dependent variable, high temp is
>independent or vice versa)
>2) take the results from step 1 and model (using proc
>mixed) against low temp
>3)using the covariance/correlation matrix generated
>from each simple regression into the equation you
>specified below (the equation with the variable
>Also, do you have an example of how the PROC MIXED
>code would look based on what you described. I should
>mention that my big economic model was estimated using
>proc mixed with an AR(1) time-series error correction
>and fixed effects for U.S. state.
>Yahoo! FareChase: Search multiple travel sites in one click.