2

I have data on spot prices, inventory, and storage capacity. I want to regress spot prices on the inventory level but the relationship appears nonlinear. I believe that the non-linearity is created when the storage utilization, i.e. inventory/storage capacity, is either very high or low. What is the correct way to capture the non-linear dynamic?

Should I regress spot on inventory + storage capacity or spot on (Inventory/Storage capacity) or something else? Do I need to use dummy variables?

Regressing spot on inventory

Steffen Moritz
  • 1,564
  • 2
  • 15
  • 22
Trendofearnings
  • 21
  • 3
  • 1
    It's not a trivial question. Using a linear spline with a knot at zero (effectively breaking the plot into right and left hand portions that meet in the middle) looks like it would accommodate the apparent nonlinearity--and might even have a meaningful interpretation in terms of economics or business practices. You don't want to apply a nonlinear transform of the spot price, tempting though that might be, because your responses look fairly homoscedastic already. – whuber Mar 29 '18 at 16:18
  • Thank you for your answer! I will look into the spline solution. If I, however, where to apply a nonlinear transformation of the spot price; how would I proceed? – Trendofearnings Mar 29 '18 at 17:28
  • I outlined one method at https://stats.stackexchange.com/questions/35711. It's inadvisable in your case, though, because you would ruin the homoscedastic responses. I suppose you could use weighted least squares to compensate, if you think a transformation would have a better interpretation than the spline. – whuber Mar 29 '18 at 18:01

0 Answers0