I'm running a pooled model sales, and some price variables are coming in +ve.
I observed the series and found that for one pool it's because the series was in decline, and despite variability indicating a -ve correlation, the decline of the series gave an overall positive correlation.
So to correct for this, I thought of including a trend variable in that pool specifically.
I've created a bunch of 0s series for the other pools, and a 0-156 series for the pool in question and entered into the model, but I'm getting a near singular matrix error. There's no reason for it to creep up apart from perfect collinearity with the constant sooo I'm not sure what's happening there.
Anyone came across this and resolved in past?
Cheers,
Elsa