OnDemand WTP Pricing Research

The Case for Logarithms | Joe Brillantes

The graphs below present the same data. The difference is the axis for quantity of the bottom graph is in natural logarithmic scale. Transforming data to their logarithms most often decreases variability, enables visualization, and linearizes the relationship between two variables. In the top graph, there seems to be three “outliers”, but when the quantity axis is logarithmically scaled, the three “outliers” are not outliers anymore. Thus, variability decreased because of the absence of outliers. It’s also easier to “see” the data in the bottom graph than in the top. A linear model of the relationship of price and log quantity is very much apparent in the bottom graph. Jeffrey Wooldridge, in his book Introductory Econometrics: A Modern Approach, suggests that logarithmic transformations are only appropriate for variables that can never be negative (because logarithms of values less than or equal to zero are not defined), and can take very large values. Furthermore, he writes that variables in percent or in proportion, and measured in time should not be logarithmically transformed.

Analyst – Statistical Analysis, Forecasting, Predictive Modeling and Optimization at Lattice Semiconductor

3 Comments on "The Case for Logarithms | Joe Brillantes"

Trackback | Comments RSS Feed

Post a Comment

WP-SpamFree by Pole Position Marketing