LINEST and Dynamic Regression
 

Hi-

I'm trying to built a multiple regression model which dynamically switches X
variables based on their independent r-squared with my y variable.

Specifically, i have 4 "potential" x variables, but I only want to use the 2
which have the strongest linear relationship with my y variable based on the
trailling 36 periods.

So for example, lets call my Y variable "Copper Price" and my X variables
"Supply", "Demand" ,"Inventories" and "USD FX". From 1995-1998, supply and
demand had the highest rsqs with copper prices, so i want to only use those.
However, by 1999, "inventories" is more important than supply, so i want my
linest function to be based on demand and inventories.

I know this is a simple example, but i've actually got a couple dozen
variables so I need a better way to do it.

Any thoughts would be very much appreciated.

-W