experimentation

Frequentist Stopping Rules

Robert Neal

Stopping rules are what make most frequentist (a.k.a. classical) statistical tests valid. Unsurprisingly, they tell you when to stop an experiment, and equally unsurprising, they are rules that must be followed in order to get statistically valid experiments from statistical tests that depend on them. However, despite stopping rules being critical to the interpretation of […]

Read More

Building Competing Models Using Apache Spark DataFrames

Abdulla Al-Qawasmeh

  Tech Talks: Building Competing Models Using Apache Spark DataFrames from Credit Karma on Vimeo. In this video, Abdulla Al-Qawasmeh, Engineering Manager, talks about the important role model calibration plays in the performance of Credit Karma’s recommendation system, and how to build predictive recommendation models using Apache Spark DataFrames.

Read More