wikimedia / research-reader-survey-analysis

Mirror of https://gerrit.wikimedia.org/g/research/reader-survey-analysis

Reader Survey Analysis

See our meta page for a more complete description of the code contained within this repository. The main goal of this code is to provide a means for debiasing survey results (to account for sample and non-response bias) based upon associated log data.

Overview

• src/preprocessing: code for cleaning survey responses and preparing them for de-biasing
  • 01_collectsurveydata: clean survey responses
  • 02_extractlogtraces: extract and preprocess log data for debiasing
  • 03_debiassurveys: build model to compute new weights for each survey response based upon the estimated likelihood that they would respond to the survey
• src/utils: various utility scripts and configuration for the code

Attribution

The code contained in this repository is largely the work of, in no order, Florian Lemmerich, Philipp Singer, and Ellery Wulczyn.