Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am running time series regressions to estimate the percentage change in quantity to a percentage change in price, with the most basic form being $\ln Q = \beta_0 + \beta_1\ln P + \varepsilon$, where $\ln Q$ and $\ln P$ are the log of quantity and price respectively. My question is if anyone knows about the validity of taking $log(q + 1) = \beta_0 + \beta_1\log(p + 1) + \varepsilon$? The reason for the transformation is obviously to deal with zeros (that is, time periods with zero sales).

share|improve this question
This question may be helpful to you: how-should-i-transform-non-negative-data-including-zeros. – gung Nov 12 '12 at 15:46

You can do this in a GLM framework with a log() link function or with a poisson regression model instead of transforming $q$. Both of these will handle the log of zero issue very well. However, getting the time series aspect of the problem right might be a challenge in this framework. For example, you might have shocks to demand that persist from day to day.

I am not sure why you need to transform $p$ since it is presumably greater than zero at all times.

Unfortunately, the elephant in the room is probably endogeneity. Unless you have a very unusual case, you are not identifying the price elasticity. I would bet dollars to donuts that your price coefficient will be positive. Take a look at these lecture notes to get an idea of what's going wrong.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.