Statistical-Inference-for-Everyone

Introductory Statistical Inference

AI & Machine Learning Science & Engineering

Explore on GitHub Read Hacker News discussions

2Distinct people

2Mentions

1Explicit recommenders

146GitHub stars

REPOSITORY AT A GLANCE

What to know before exploring

What it is

Introductory Statistical Inference

GitHub

Where it is used

No documented product use has been verified yet.

What people are saying

> But simply sampling 10 people is about as good as 10 anecdotes. That is not generally true. It depends on what kind of statistical assumptions you make, or statistical…

Pyxl101 used a recommendation phrase. One indexed comment matches. This is a phrase match, not a verdict.

No classified warning is attached to this page. Read the discussions before treating silence as approval.

Also in these discussions

No other repository is repeatedly linked in the same discussions.

This week on Hacker News

Nov 7–Nov 14 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/14/2026, 3:56:07 PM UTC.

No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.

Mentions over time

HN mentions by week2 dated mentions

Observed mentions: 2016-12-22 05:18:18 UTC to 2017-01-26 10:42:43 UTC. Weeks start Monday, UTC.

Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.

View mention counts and date ranges

UTC ranges include their start and exclude their end. Select a period to filter the source evidence.
Period	Start (UTC)	End (exclusive, UTC)	Mentions	Comments
Dec 19	2016-12-19 00:00:00 UTC	2016-12-26 00:00:00 UTC	1	1
Dec 26	2016-12-26 00:00:00 UTC	2017-01-02 00:00:00 UTC	0	0
Jan 2	2017-01-02 00:00:00 UTC	2017-01-09 00:00:00 UTC	0	0
Jan 9	2017-01-09 00:00:00 UTC	2017-01-16 00:00:00 UTC	0	0
Jan 16	2017-01-16 00:00:00 UTC	2017-01-23 00:00:00 UTC	0	0
Jan 23	2017-01-23 00:00:00 UTC	2017-01-30 00:00:00 UTC	1	1

What Hacker News says

2 sources

MLWiDA·Jan 26, 2017

I like to throw in the first chapter of "Statistical Inference for everyone" for a little bit more explanation for these rules (https://github.com/bblais/Statistical-Inference-for-Everyone). It looks to me like an easier approach to the heavy weight "Probability Theory: The Logic of Science" by E.T. Jaynes

In A Cheat Sheet on Probability

Pyxl101·Dec 22, 2016Recommendation phrase

> But simply sampling 10 people is about as good as 10 anecdotes. That is not generally true. It depends on what kind of statistical assumptions you make, or statistical analysis you conduct. The term statistical power describes our chance of correctly detecting an effect when there is one to observe. If the effect is small, then a larger number of samples are required to achieve a given level of statistical power, whereas if the effect is large, then fewer are required. I recommend that you consider the question: if 10 samples are not enough, then what specific number of samples is enough? How do you decide? Fortunately, these questions have been studied in the field of statistics. Let me give you an example. Let's say I told you that you will flip a coin in the air, and when the coin reaches its peak height, I will shout "heads" or "tai

Read the rest

ls". What would you make of it if we ran this experiment 10 times, and I correctly guessed the outcome 10 out of 10 times? Perhaps you would conclude I really can predict coin flips. By comparison, if I called the outcome accurately only 5 out of 10 times, then you'd probably consider my claim false. But, consider these two possibilities: (1) what are the odds that my guesses are really no better than random chance, and I've just guessed 10 out of 10 correctly by good luck? (2) What if I really am accurate 99% of the time, but I guessed only 5 out of 10 correctly by bad luck? Statistics allows us to evaluate how likely these things are. If I'm doing the math correctly, then you'd expect someone to guess 10 out of 10 coins correctly just by chance once in every ~1000 experiments. So to see this happen is not witnessing an extraordinarily improbable event; run enough experiments of 10 coin flips and you will see it. If someone truly has 99% accuracy, then in almost every experiment they will guess 10 out of 10 correctly. They should guess all 10 correctly 90% of the time. A person who is truly 99% accurate will only guess 5 out of 10 flips correctly once every in every 10,000,000,000 experiments. So it is extraordinarily unlikely that you will see someone with 99% accuracy guessing 5 out of 10 coin flips correctly. It can still happen just by chance, but it's really improbable. Bringing this all back to the main topic, it is possible for a result of 10 data points to count as convincing evidence against the theory that there is a strong effect, such as that musicians are 99% accurate in discerning the type of violin, while it may be inadequate to evaluate whether there is a weak but still-present effect, such as 51% accuracy. Whether the number of samples is good enough depends on how small of an effect you want to measure, and how confident you want to be in your assessment. A free book is available online called "Statistical Inference for Everyone" which introduces these topics. https://github.com/bblais/Statistical-Inference-for-Everyone

In The Brilliance of a Stradivari Violin Might Rest Within Its Wood