Posts tagged statsmodels

Statistical Models from formulas

This week, I taught a course on statistical modeling in statsmodels. For those of you who have never used or heard of this Python package, it began as a subpackage in scipy called scipy.models. However, as it grew in size and complexity, it was removed from scipy, and then it became its own package, statsmodels.

As a package, it is a great way to carry out statistical modeling as it provides a great deal of model introspection right out of the box, enabling users to fine-tune their model specification. In this regard, it is similar to the very popular scikit-learn package, but I have found the main difference between the two is that statsmodels is more for introspecting single models, while scikit-learn provides a powerful, object-oriented interface for creating predictive pipelines.

Read more ...