Embedded Linux Quality Assurance: How to Not Lie with Statistics
summary
Argues that statistical QA is essential for embedded Linux reliability and latency claims and demonstrates concrete tools and steps: using R and the grammar of graphics on tidy data, plotting faceted histograms and an empirical CDF with a square-root transform to expose spikes around ~55 μs and ~100 μs in a ~60k-sample multi‑core latency run. Shows mean+std is misleading, emphasizes reporting sample size, and gives a cautionary example where linear regression on bug‑fix counts violates model assumptions and yields unsafe predictions.
AI-written summary · about
chapters (24)
- 0:00Introduction
- 2:04How to Not Lie with Statistics
- 4:33Testing
- 5:35Certification
- 7:24Why Statistics
- 8:20Dealing with Data
- 8:36Measuring Data
- 11:39Tidy Data
- 14:01Data in messy form
- 16:30A proper statistical environment
- 17:59How to understand data
- 19:28Types of data
- 22:44Histogram
- 27:39Transformations
- 29:44Nonparametric method
- 31:56Summary statistics
- 34:07Summary statistics example
- 35:09Comparing data sets
- 37:45Cumulative distributions
- 38:36Common mistakes
- 42:43Linear Regression
- 45:00Linear Regression Results
- 47:00But Fix Commits
- 50:35Summary
related talks
40mReal Quality Assurance from an Agile and Lean Viewpoint
32mStatistics for Engineers
41mLean QA: Much more cost-effective Quality Assurance methods, than testing
29mYour Metrics Are Lying: How to Stop Reporting and Start Learning
41mA Human-Centric Quality Assurance Process for Open Source Softwa...
15mLinux Testing made better with DATA
38mDebugging Embedded Linux
36mLina Zubyte: Escaping the Trap of Infinite Testing and the Superiority Complex of a QA