HN's go-to recommendation for colormap and color-perception questions; commenters call it funny, brief, clear and a favorite talk.
Topics › AI, Machine Learning & Data
The best data science & statistics videos, according to Hacker News
55 videos about data science & statistics that Hacker News commenters keep pointing each other to, ranked by how many separate months they were recommended. Part of AI, Machine Learning & Data.
HN repeatedly recommends it to counter overpopulation alarm with data on falling birth rates.
HN commenters link it whenever a shaky study or the p<0.05 threshold comes up, as a short primer on false positives in science.
Recommended as evidence that the world is improving despite negative news, and for how people misjudge global trends.
It is the go-to funny, irreverent critique of notebooks, linked whenever HN debates notebooks versus testable code.
HN recommends it as the fastest way to understand what VisiData can do, since seeing it work is clearer than reading about it.
Repeatedly recommended as the canonical demo of org-mode for literate programming and reproducible research.
Whenever notebooks get criticized, HN commenters point to it as the standard rebuttal from a prominent practitioner.
HN recommends it as an enthusiastic demo of spreadsheet-like reactive notebooks, which commenters think new notebook projects should copy.
HN commenters call it an internet classic that makes you a more proficient Excel user, and note it is still relevant despite predating newer functions like XLOOKUP.
HN points to it as a thoughtful, first-principles take from a Turing award winner that cuts through NoSQL and big data hype.
HN repeatedly recommends this series as a clear, step-by-step way to really understand Kalman filtering.
HN commenters point to it as a first-hand overview of Heap's Postgres/Citus architecture and its distinctive indexing approach.
HN commenters recommend it as a data-driven explanation of the 2024 election and voter trends.
Cited in threads about Python's rise as a first-hand, entertaining account of why it won in scientific computing.
HN brings it up in any thread about German or European rail as a data-driven, entertaining analysis of why trains are late.
HN commenters bring it up whenever a poll or survey result looks suspect, as the classic illustration of biased question framing.
HN commenters recommend it for the finding that idea exchange and connectivity drive productivity and wealth, which they apply to cities, offices and online communities.
Cited as the classic data-driven argument that the first-world versus third-world split is outdated and the world is improving.
HN commenters recommend it as a technically jaw-dropping talk, and it is the standard reference whenever the Night Watch gigapixel images come up.
HN commenters link it as the clearest short explanation of regression to the mean, especially for how it misleads people about feedback and statistics.
HN recommends it as an excellent programmer-friendly explanation of p-values and statistical thinking.
HN recommends it as the definitive talk from the implementer on DuckDB's motivation and design.
Cited in threads on YouTube algorithms and creator complaints as a data-driven look at the trending tab.
Recommended as a firsthand history of how Python became the language of scientific computing and ML.
HN points to it as the company's own pitch, used in debates about data-driven manipulation and extrapolated to today's AI.
HN recommends it as the clear explanation of why Google's statistical, data-first approach works.
HN people return to it as a clear primer on MCMC and HMC and on why high dimensions behave counterintuitively.
Recommended as a favorite tech talk showing how huge data availability changes how problems are approached.
HN commenters recommend it for changing how they think about distributions and for showing how criticality shows up across earthquakes, wildfires and other fields.
HN points to it as the best explanation of why R is so hard to optimize, often paired with talks on Python and used to explain what makes Julia fast.
It is recommended to programmers who want to learn statistics through simulation rather than memorized formulas.
Recommended as a vivid visual demonstration of how unreliable p-values are.
HN commenters point newcomers to it as a solid starting point for getting to know Julia.
Recommended in threads on German rail as the data-backed explanation of how reported punctuality understates real delays.
Recommended as a short, fun introduction to micromorts and thinking about risk.
HN points to it as an accessible case that diets work differently for different people, partly due to gut bacteria.
HN cites it when arguing that different risks follow different distributions and should be judged by growth rates and tails rather than current counts.
HN points to it as a clear explanation of what flood return periods really mean.
HN points to it as an accessible introduction to Bayesian reasoning and as inspiration for personal projects.
HN links it in data-visualization threads as an entertaining case against violin plots.
HN commenters cite it as a clear debunking of election-fraud statistics and as a warning against comparing percentages without context.
Commenters cite it as a visual debunking of Benford-based election fraud claims and as a guide to when Benford's Law does and does not apply.
Linked as the clearest technical explainer of differential privacy and the census, and of why aggregate data can still compromise privacy.
Used to illustrate that humans pattern-match and find spurious patterns in enough data points.
Cited again and again as the concrete case study behind the claim that NASA got a 15,000x speedup moving from Simulink to Julia.
Pointed to as the clearest explanation of what Foundry actually is, and for its polished software, though one commenter calls it dated now.
HN uses it as a handy reference whenever a thread needs to point at survivorship bias.
HN people link it as a short, accessible summary of Judea Pearl-style causal inference when someone repeats that correlation cannot imply causation.
It is mostly cited as a joke or a wry reference in threads about dubious statistics.
HN cites it as a classic illustration of dysfunctional management, gamed targets, and blaming workers for systemic outcomes.
Cited in modeling and machine learning threads as the classic story on overfitting and justifying parameters.
Cited as a mathematician's debunking of numerical claims of 2020 election fraud.
Linked whenever surveys or polls are suspected of manipulation through question wording.
Invoked as shorthand for hiring and investing by gut feel and appearances instead of what actually predicts results.
Which of these have you already seen?
Connect YouTube and HackerTube AI hides what you've watched and ranks the rest for you.