Finding Firefox Beta Subsets Resembling Release

Work related to the application of Statistical matching methods to finding subsets of Firefox Beta users that are representative of Release. These subsets can be utilized to forecast Release behavior before it is launched to the populace.

Project

The primary project details are given in the PRD. The results and deliverables of the project milestones are contained in the perf_release_criteria directory.

Proof-of-concept

Initial proof-of-concept work is contained in the poc directory. The MatchIt library in R was used for matching.

• data_prep: contains pyspark scripts (converted from Databricks notebooks) that created datasets from Firefox telemetry pipeline.
  • Beta_Release_Matching_Perf_Metrics_POC.py: data munger for the initial analysis.
  • Beta_Release_Matching_Perf_Metrics_Validation.py: data munger for the validating analysis
• analysis: contains R Markdown, scripts, and final rendered html reports.
  • modeling_poc.Rmd: grunt exploratory work regarding initial statistical modeling efforts.
  • report_poc.Rmd: final R Markdown report regarding initial statistical modeling.
  • report_validation.Rmd: final R Markdown report regarding method validation.