Major·Fluent

№ 24 · 8 modules × roughly 3 hours

Statistics

Every number you trust was built on a sample — learn to know when to trust it.

In 24 hours you will become conversant in Statistics' vocabulary, core frameworks, and field debates — able to read a study critically, reason about data under uncertainty, and recognize when numbers are being used to mislead. This builds genuine fluency, not a statistician's credentials.

Time
8 modules × roughly 3 hours
Difficulty
Introductory but serious
Adjacent fields
Data Science · Economics · Psychology · Biology / Pre-Med · Computer Science

Contents

8 modules · ~3h each · ~24h total
01
Field Orientation

Orient the learner to what statistics actually is — learning from data under uncertainty — and establish the honest scope of this course: statistical fluency for critical reading, not credentials, not a substitute for real mathematical training or a professional statistician.

02
Vocabulary Immersion

Build the working vocabulary of statistics — across description, probability, sampling and inference, and relationships — so you can read a study, parse a headline, and name exactly what is wrong when a number is being misused.

03
Mental Models

Build the reflexes — not frameworks to apply consciously, but fast intuitions that fire automatically when you encounter a number — so that your first instinct about any statistical claim is to ask the right questions before you believe or dismiss it.

04
Frameworks and Theories

Equip learners with the six core statistical frameworks — probability and distributions, the Central Limit Theorem, estimation and confidence intervals, null-hypothesis significance testing, correlation and regression, and frequentist vs Bayesian inference — understood as analytical lenses with real explanatory power and honest limits, not plug-in formulas.

05
Methods and Tools

Build working fluency in how statistical studies are actually conducted — from data collection through inference — and develop the critical habits to spot the ways statistics fools even careful researchers: p-hacking, multiple comparisons, publication bias, and the garden of forking paths.

06
Canonical Cases and Debates

Ground statistical ideas in the real thinkers who built them and the real cases that tested them — from Fisher's teacup to the replication crisis — so the abstractions stick and the field's genuine debates feel live rather than academic.

07
Applied Project Studio

Build a Statistical Critique Memo — a structured, honest appraisal of a real data-based claim — working through six staged steps: stating the claim precisely, auditing sample and design, assessing uncertainty, examining the causal warrant, hunting for signs of self-deception, and reaching a calibrated verdict.

08
Synthesis and Fit

Integrate every reflex and framework from the sprint into one coherent map, name the Dunning-Kruger cliff honestly, assess whether statistics is a field worth pursuing further, and leave with a 30-day plan for keeping fluency alive.

After this sprint, you can…

Fluency, not mastery
  • Use Statistics's core vocabulary without bluffing.
  • Recognize the field’s major debates and the tradeoffs behind them.
  • Ask sharper questions of practitioners, books, courses, and AI tools.
  • Read entry-level sources with enough context to judge them.
  • Spot common beginner overclaims — including ones an AI might make.
  • Decide whether deeper study, expert help, or formal training is worth it.

Canonical frameworks

  • Descriptive vs inferential statistics: descriptive methods summarize the data you have; inferential methods use that data to draw conclusions about a larger population with quantified uncertainty
  • Probability and probability distributions: probability is the language of uncertainty; distributions (normal, binomial, and others) describe how values are spread and make probabilistic calculations possible
  • Sampling distributions, the Central Limit Theorem, and the Law of Large Numbers: these theorems explain why sample means behave predictably and why normal-distribution methods apply so broadly in inference
  • Estimation and confidence intervals: point estimates summarize sample data; confidence intervals honestly quantify the range of plausible parameter values and the uncertainty inherent in any sample
  • Null-hypothesis significance testing (and its errors and power): the formal framework for asking 'could this result be noise?' — with its celebrated p-value, its type I and II errors, power, and its deep and ongoing controversies
  • Correlation and regression: tools for modeling relationships between variables — with the essential caveat that association is not causation and that confounders lurk in every observational dataset
  • Frequentist vs Bayesian inference: two philosophically distinct answers to 'what does probability mean?' — long-run frequencies vs degrees of belief — with different tools and different interpretations of the same data
  • Experimental design (randomization and control): how data is collected determines what conclusions it can support; randomization is the only reliable path to causal inference, and its absence is the root of most observational confounding

Live debates

  • Frequentist or Bayesian: what does probability actually mean?
    The debate determines how you interpret every p-value, confidence interval, and statistical test — and whether you can ever make a probability statement about a specific hypothesis being true, which most scientists desperately want to do but frequentist statistics does not technically permit.
  • Is the 0.05 p-value threshold defensible, and should we abandon or redefine 'statistical significance'?
    The 0.05 threshold is the de facto gatekeeper for scientific publication in many fields — a single arbitrary number determines what gets published, funded, and acted upon, with enormous downstream consequences for what the public and policymakers believe is true.
  • Statistical significance vs practical significance: is a p-value alone enough to report?
    Vast resources — drugs approved, policies enacted, products launched — have been justified by 'statistical significance' without any examination of whether the effect was large enough to matter, producing real-world decisions grounded in noise amplified by large samples.
  • What caused the replication crisis, and how do we fix it?
    The replication crisis called into question large swathes of published psychology, medicine, nutrition, and social science research — findings that had shaped clinical guidelines, public health campaigns, and policy were found to be fragile or false, with real costs to public trust in science.
  • When, if ever, can observational data establish causation?
    Most data in medicine, economics, and social science is observational — if observational data can never establish causation, vast domains of human knowledge are fundamentally limited; if it sometimes can, the criteria for 'when' become the most important question in applied science.

Source trail

7 notes
  1. Diez, Cetinkaya-Rundel, and Barr, OpenIntro Statistics (4th ed., open access) — the gold-standard free introductory text; used for core inference and probability framing
  2. David Spiegelhalter, The Art of Statistics: Learning from Data (2019) — authoritative, accessible treatment of modern statistical reasoning; backbone for the replication crisis and uncertainty modules
  3. Darrell Huff, How to Lie with Statistics (1954) — foundational catalog of statistical deception tactics; still the best short primer on misleading data presentation
  4. Charles Wheelan, Naked Statistics: Stripping the Dread from the Data (2013) — narrative-driven conceptual coverage; used for worked examples and accessible framing of inference
  5. John Ioannidis, 'Why Most Published Research Findings Are False,' PLoS Medicine 2(8), 2005 — the replication crisis anchor; canonical for p-hacking, publication bias, and the limits of significance testing
  6. American Statistical Association, 'Statement on Statistical Significance and P-Values,' The American Statistician 70(2), 2016 — the profession's own reckoning with p-value misuse; essential for the significance debate
  7. Daniel Kahneman, Thinking, Fast and Slow (2011) — base rates, representativeness, and cognitive biases that corrupt statistical reasoning; grounds the mental-models module

Watch the field

3 curated videos · included

This field includes a curated shelf of 3 hand-picked free explainer videos — vetted from trusted educators and embedded so you can watch them in context, without falling down the YouTube rabbit hole. A small bonus on top of the eight-module sprint; it unlocks with the field.

Ask better questions of AI

Fluency is leverage

Fluency in Statistics makes AI far more useful: you know what to ask, you can judge the answer, and you know when to check a primary source or a practitioner instead. Once you've done this sprint, prompts like these get real work done — using the field's own frameworks and debates:

  • I'm new to Statistics. Define <term> the way a practitioner would, give one realistic example, and flag where beginners misuse it.
  • Apply Descriptive vs inferential statistics: descriptive methods summarize the data you have; inferential methods use that data to draw conclusions about a larger population with quantified uncertainty to <my situation> and show your reasoning — then list what could make this analysis wrong.
  • Lay out both sides of: Frequentist or Bayesian: what does probability actually mean? Give the strongest evidence for each, and say where practitioners still disagree.
  • Critique my plan using Probability and probability distributions: probability is the language of uncertainty; distributions (normal, binomial, and others) describe how values are spread and make probabilistic calculations possible. What assumptions would a Statistics practitioner question?
  • What primary sources or practitioners should I check before trusting your answer on <topic> in Statistics?

Expert · AI · Source. Use AI to orient and draft, primary sources to verify claims that matter, and a practitioner when judgment, liability, or nuance is on the line. Fluency is what lets you tell which is which.

What this sprint does not do

This is field fluency, not mastery — and not credit, licensure, or professional authority. It does not qualify you to practice Statistics where supervision, certification, or a license is required. It gives you the operating language and judgment to learn faster, ask better questions, work with AI and experts, and decide your next move.