Python Data Analysis Cookbook

Applying logit() to transform proportions

We can transform proportions or ratios with the SciPy logit() function. The result should be a more Gaussian distribution. This function is defined by the following equation:

As you can see in equation (4.4), the logit is the logarithm of the odds. What we want to achieve with this transformation is getting a more symmetric distribution—a skew close to zero. As the proportions approach zero and one, the logit asymptotically approaches minus infinity and infinity, so we have to be careful in those cases.

As an example of a proportion, we will take the monthly proportions of rainy days. We get these proportions ...