Skip to content

Experiments showing the profit efficiency of targeted randomized sampling in comparison to standard A/B testing

License

Notifications You must be signed in to change notification settings

johaupt/supervised_randomization

Repository files navigation

Supervised Randomization

Code accompanying the paper: Haupt, Jacob, Gubela & Lessmann (2019). "Affordable Uplift: Supervised Randomization in Controlled Experiements".

Summary

Supervised randomization integrates the targeting model into randomized controlled trials and A/B tests as a stochastic policy. By replacing random targeting with supervised randomization, we reduce the costs of running randomized experiments or allow continuous collection of randomized trial data. We show how to fully correct downstream analysis for the bias effect by the supervised treatment allocation with the true probabilites to receive treatment, which are logged at the time of targeting.

Data

The data we use for empirical monte carlo simulation is public at https://archive.ics.uci.edu/ml/datasets/Bank%2BMarketing

Code

evaluation_script.R contains the code to replicate the tables and figures presented in the paper load_bankmarketing.R contains the code to preprocess the bank marketing dataset

About

Experiments showing the profit efficiency of targeted randomized sampling in comparison to standard A/B testing

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages