1

I have to estimate a model forecasting the sales (as stock units) for AXE deoderants.I want to apply the multiplier specification on this model. The model should look like this:

log(Sales) = b0 + b1*log(Price) +b2*log(Adertising) + log(b3)*Summer

The last variable is causing the trouble. It is a dummy variable that is set equal to 1 for sales during the summer and 0 elsewhere. I have no idea how a logarithm of a parameter can be taken in R.

Can somebody please help me to model this?

Many thanks in advance!

whuber
  • 322,774
Pieter
  • 13
  • 2
    This is a linear model (if you substitute $\log(b_3) = b_3'$). Function lm can deal with factor variables and even does the dummy encoding for you. – Roland Oct 22 '14 at 08:49
  • Well, if you exponentiate it you get the original multiplicative model. It is not linear in the parameters. I'm not sure if I can do this in R. The parameter (b3) is estimated automatically if I run the regression. It is not clear to me how I force the software to take the logarithm of it. – Pieter Oct 22 '14 at 09:22
  • You have to decide whether you want to fit the non-linear model or the linearized model (in your question you express a preference for the latter). Those are not equivalent. – Roland Oct 22 '14 at 10:46
  • You've written log(b3). Do you mean log(Summer)? – Sycorax Apr 13 '15 at 20:18

0 Answers0