3

I have been through many posts arguing about whether we should use the natural log of a ratio or not. Yet I am still unable to understand the concept completely for I feel even ratios differ from each other like child-teacher ratio is not like the dependents to working age population ratio. How do we evaluate it?

To be very specific in one of the old posts titled: In linear regression, when is it appropriate to use the log of an independent variable instead of the actual values?, it has been posted with reference to a book that it is better to use proportions in their level form instead of transforming them into their natural log. As I am quite clear about the interpretations when logging either one or both sides of a regression equation, I am confused if convenience of interpretation is the only reason not to choose log transformation of a ratio or proportion?

Community
  • 1
Nida Jebran
  • 31
  • 1
  • 5
  • 4
    Ratios often behave better when logged, so long as both numerator and denominator are positive, and their analysis is often simpler. If either numerator or denominator is zero, then what to do depends on the details. I don't understand your uneasiness here; naturally ratios can differ from each other, but so can anything else. – Nick Cox Mar 04 '16 at 14:45
  • Could you please provide a link to the "old post" you are referring to? Googling turns up the thread at http://stats.stackexchange.com/questions/298, but I am having trouble finding any answer that says what you claim. Moreover, I find a large number of answers providing a broad spectrum of useful advice, all of which appears to contradict your assumption that the only consideration is "convenience of interpretation." – whuber Mar 04 '16 at 18:51
  • 1
    http://stats.stackexchange.com/a/177624/107465 – Nida Jebran Mar 05 '16 at 12:36

0 Answers0