Tabs, spaces and salaries: a data science detective story
summary
Stack Overflow’s claim that developers who use spaces earn more is reproduced from the survey but investigation shows why: a regressed model (full vs reduced) and linear regression with adjusted R^2 find indentation adds no predictive power once other variables are present. A mixture model for monthly vs annual uncovers many countries reporting monthly pay as annual, and a chi-square on version control vs salary plus splits by open-source contribution, language and country reveal company type and tooling (git, OSS-heavy languages) drive the gap. All analysis is reproducible on GitHub.
AI-written summary · about
chapters (7)
- 0:00Start
- 21:57COLLINEARITY?
- 26:21LANGUAGE AND OPEN SOURCE?
- 36:01WHY IS VERSION CONTROL SO IMPORTANT?
- 47:18Voter Embarrassment about Trump Support May Have Messed Up Poll Predictions
- 48:42CAN YOU TRUST DATA?
- 51:44EVELINA GABASOVA Consulting data detective
related talks
45mTabs, spaces and salaries: a data science detective story
46mTabs, spaces and salaries: a data science detective story
01h 04mWhat can we learn from 750 billion GitHub events and 42 TB of code
36mA developer's guide to probabilistic programming
18mData Analysis with JavaScript
55mExploring StackOverflow data
01h 03mWhat can we learn from 750 billion GitHub events and 42 TB of code
20mIT Holy Wars