1

I have a regression model where the dependent variable is logarithm of miles per gallon and several IVs. One of the IVs is origin of the car where 1=USA, 2=Europe, 3=Japan.

I wonder how to interpret the results of the model if the beta parameter for the variable origin is 0.040.

Furthermore isn't it better to transform the variable origin to 3 dummy variables USA, Europe and Japan?

The dataset comes from here.

Nick Cox
  • 48,377
  • 8
  • 110
  • 156
Adasz
  • 105
  • 1
  • 1
  • 9
  • 2
    You should indeed use indicator or dummy variables for origin coded 1, 2, 3. The codes are quite arbitrary and the information should be represented using indicator variables. It's far from obvious that miles per gallon should be logged.... – Nick Cox May 02 '15 at 12:11
  • Sorry but I don't follow you :/ So I should transform the discrete variable to three dummy ones? What do you mean by miles per gallon should be logged? – Adasz May 02 '15 at 12:16
  • Yes; you should use indicator variables, as stated. I am questioning why you are logging miles per gallon, not recommending it. – Nick Cox May 02 '15 at 12:33
  • Now I see. I have transformed the mpg because of heteroskedasticity in the data. – Adasz May 02 '15 at 12:35
  • A better motive would be getting closer to linearity. Working with gallons/mile often helps as this is closer to underlying physics. – Nick Cox May 02 '15 at 12:37
  • @Adasz Take a look at the first part of [this answer](http://stats.stackexchange.com/a/93103/7071). – dimitriy May 04 '15 at 17:50

0 Answers0