← all systems · see it in The System
Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on
Statistics is the discipline that concerns the collection, organization, analysis, interpretation, and presentation of data. ... Statistics is considered the science of uncertainty. ... The academic discipline of statistics formalizes these intuitive statistical inferences.
Properties published scienceour proposal
Sourced fields, and our own readings.
| born at | not dated for the intuitive floor (birds and primates, convergently); the discipline: Al-Kindi (9th century, as reported), Graunt 1662 |
| born precision | undated for the sense; year for Graunt |
| boundary | Inside: the practice of statistics (collecting data by census, survey, experiment and records; describing it; inferring from samples to populations; deciding under uncertainty), the people, offices, tools and computers that do it, the intuitive statistics in brains that the discipline 'formalizes' (statistical learning, ensemble summaries, inference from samples), and the structure it works on (samples, distributions, statistics, estimators, tests). Outside: the populations and processes measured (people, atoms, telephone amplifiers, patients), the probability theory it rests on (its own walk), and the states, sciences, firms and clinics that act on its conclusions. |
| nested level | Sensed in vertebrate brains (rung 8 up, convergent in birds and primates); practised and written by people at rung 11; run on computers at rung 12; its structure is the probability walk's (roles). |
| scale | social |
Measures (20) published science
| characteristic clock: exposure needed for 8-month-olds to find words in a syllable stream | 2 minutes |
| transitional probability within words / across word boundaries (the stream's statistic) | 1.0 / 0.33 probability |
| age at which infants infer populations from samples | 8 months |
| kea first-trial accuracy across all conditions | 72.22 % of trials |
| kea tested | 6 birds |
| time since birds and humans shared an ancestor (bound for convergent statistical inference) | at least 312 million years |
| Graunt's funerals in London | about 13,000 per year |
| London population estimated by Graunt (first ratio estimator) | about 384,000 people |
| birth year of statistics as usually dated (Graunt) | 1662 year |
| term statistics in modern use (Achenwall) | 1749 year |
| census period recommended by the UN | at least every 10 years |
| PPV of an adequately powered RCT with 50% prior | about 85 % |
| conventional significance threshold | 0.05 p |
| conventional confidence level | 95 % |
| Bayesian answers among 160 gynecologists: probabilities vs natural frequencies (per-hand-off efficiency of the format) | 21 vs 87 % correct |
| gain from natural frequencies, medical students, four complex tasks | 37 percentage points |
| Bayes answer to the mammography problem / physicians' typical estimate | 7.8 / 70-80 % |
| students with some statistics anxiety | more than 70 % |
| information kept by a sufficient statistic (hand-off efficiency, upper bound for any summary) | all (η = 1); any post-processing η ≤ 1 share of information about the parameter |
| brain's share of body energy (the cost critics weigh against exact Bayesian inference) | about 20 (2% of mass) % of body energy |
Open contentions (4) open contention
Where sources or readings disagree. Both stay on the map.
- C-980 Does the brain do Bayesian statistics?
booked: The map records that the brain tracks frequencies and patterns, and leaves open whether it does true Bayesian inference.
the other reading: Bayesian brain / predictive coding: the brain as a prediction machine using Bayesian inference (as summarised by Mangalam 2025); critique: 'between useful metaphor and legitimate mechanistic explanation'; heuristic, non-Bayesian mechanisms (Mangalam 2025); memory-based accounts: statistical patterns as a result of the structure of memory (Saffran & Kirkham 2018) - C-981 Can we sense the average of a group of things without paying attention?
booked: The map follows the 2025 finding that the average forms only with attention, and keeps the earlier findings that it needs little or none alongside.
the other reading: earlier studies: ensemble perception without or with minimal attention (as listed by Kimchi & Sabary 2025) - C-983 Which approach to statistics is right: frequentist or Bayesian?
booked: The map keeps both approaches as variants and uses both for drawing inferences.
the other reading: frequentist: tests, confidence intervals, error rates over repeated sampling; Bayesian: prior to posterior; needs a prior; Deming: the control chart is a heuristic, 'not a hypothesis test' - C-984 Should science keep using p < 0.05 as the test of significance?
booked: The map records p < 0.05 as the convention in use, along with its criticism, and doesn't rule on keeping, changing or dropping it.
the other reading: Ioannidis 2005: single studies at p < 0.05 make most findings false in many fields; McShane & Gal (as discussed by Laber & Shedden 2017): errors cluster at the threshold (dichotomisation); Laber & Shedden 2017: decisions stay binary, with thresholds adapted to the problem
Cross-scale patterns seen here (8) our proposal
Shapes that recur when the same kind of problem is solved at different scales, and how each shows in Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on. Each links to every other system that shows it.
- same role, different builds (structural)
Statistical inference in kea and great apes with brains 'of a highly different structure': 'may have arisen twice on our planet via convergent evolution'. - same shape, opposite sign (structural)
Probability and statistics run the same mapping in opposite directions: probability deduces samples from given population parameters; inference 'moves in the opposite direction'. - The carry (structural)
The wait and the lookahead. A fixed-sample test waits until all n are in; 'traditional statistical analysis methods ... require aggregation of measurements over time, which can delay decision making' (the wait). Sequential analysis is the lookahead: data 'evaluated as it is collected', stopped by ru - The Payment to Entropy (structural)
Signal form. Every hand-off from data to summary to conclusion can only keep or lose information about the population ('post-processing cannot increase information'); only a sufficient statistic passes the parameter's information on whole (η = 1). Format is a toll too: the same screening evidence is - trade one property for another (structural)
'a computational trade-off between learning the specifics of individual experiences and regularities that hold across those experiences', met in the hippocampus model by two pathways (statistics vs episodes). And the old trade inside testing: Type I against Type II error. - Umwelt: each system carves its own slice of one shared medium (structural)
Same tones, different statistics: 'Both groups of participants in these studies heard the same sequence of tones in the input, but the groups appear to have learned different things because they tracked different primitives' (infants absolute pitch, adults relative pitch). And kea and apes take rela - templating (partial)
1853: a congress 'to unify measurement in statistical research'; UN definitions for censuses worldwide. - the whole sets the local reference (partial)
A draw is judged against its own population: true statistical inference uses 'the relative numbers of items within and between populations', so 20 good tokens mean little in a jar of 100 bad and much in a jar of 4.
Valence: what it trades, and with whom (26) published science
Each exchange with a neighbour, read from both sides: who gains, who pays, and what carries it.
- Vector space: sets whose elements can be added and scaled, as mathematics describes it, with bases and dimension, linear maps, inner products, and their use as the state spaces of physics and data via sys:computer-memory · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Hamiltonian and Lagrangian mechanics (analytical mechanics: classical mechanics built from energy functions and the stationary action, and the practice of using them) via sys:computer · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Statistical mechanics via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Election via sys:human-society · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Human (Homo sapiens) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Science (as a human social and cultural system) via sys:written-record · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Government via sys:human-society · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Hippocampus (memory) via sys:electrochemical-gradient · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Measurement (the practice of comparing a quantity with a unit, and the units and standards it uses) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Computer via sys:written-record · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Kea (Nestor notabilis) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Probability: weighing chances, in brains and in people's mathematics, and the structure it works on via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Health care via sys:human-society · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Linear algebra (the practice of solving, transforming and taking apart linear systems with vectors and matrices, and the structures it works with) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Insurance via sys:mental-representation · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Accounting (the field and practice) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Physics (the field) via sys:mental-representation · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on mixed
- Experiment, measurement and uncertainty (the physics lab as a practice: designing, measuring, estimating error) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked via sys:written-record · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- World population (human population) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Biology (the field) via sys:written-record · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Economics (the field) via sys:written-record · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on even
- Linguistics (the field: the practice and the languages it studies) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on even
- Nursing (the field: the practice and the care it gives) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Combinatorics: the practice of counting, arranging and choosing finite structures (permutations, combinations, partitions, designs, graphs), the structures it studies, and the people and machines who work with them via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
- Inflation (economics) via sys:sensory-environment · Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on gain
Partners and parts published science
Built from / builds (9)
- Hippocampus (memory)
- Intuitive statistics
- Statistical learning
- Ensemble perception
- Descriptive statistics
- Statistical inference
- Sample (statistics)
- Sufficient statistic
- Control chart
Sits inside / holds (3)
- Science (as a human social and cultural system)
- Government
- Human society (from local community up to societies)
expressed by (16)
- Templating / Standardization
- Interoperability/Transduction (Bridging)
- Encapsulation & Modularity (Abstraction)
- Sensing/Scanning
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Active Experimentation / Probing (Optimal Experimental Design)
- Multi-Scale Data Fusion / Triangulation
- Trail-Marking / Externalized Memory (Stigmergy)
- Surprise / Change-Point Detection
- Hypothesis Portfolio / Multiple Hypothesis Tracking
- Optimal Stopping & Patch-Switching
- Social Learning / Public Information Use
- Self-Monitoring
- Validation of Effectiveness
- Quality Control & Error Correction
floor filled by (36)
- Vector space: sets whose elements can be added and scaled, as mathematics describes it, with bases and dimension, linear maps, inner products, and their use as the state spaces of physics and data
- Large Hadron Collider (LHC)
- Election
- Science (as a human social and cultural system)
- Government
- Written record (durable external marks: texts, archives, canon)
- Human society (from local community up to societies)
- The world through the senses (sensory environment)
- Hippocampus (memory)
- Measurement (the practice of comparing a quantity with a unit, and the units and standards it uses)
- Computer
- Probability: weighing chances, in brains and in people's mathematics, and the structure it works on
- Intuitive statistics
- Ensemble perception
- Descriptive statistics
- Statistical inference
- Sample (statistics)
- Control chart
- Census
- Health care
- Linear algebra (the practice of solving, transforming and taking apart linear systems with vectors and matrices, and the structures it works with)
- Accounting (the field and practice)
- Psychology (the mind and behaviour as psychology describes them)
- Physics (the field)
- Experiment, measurement and uncertainty (the physics lab as a practice: designing, measuring, estimating error)
- Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked
- Risk management
- Chaos and nonlinear dynamics (the practice, and the chaotic behaviour it studies)
- World population (human population)
- Biology (the field)
- Economics (the field)
- Political science (the field: the practice and the politics it studies)
- Linguistics (the field: the practice and the languages it studies)
- Environmental science (the field: the practice and the environment it studies)
- Combinatorics: the practice of counting, arranging and choosing finite structures (permutations, combinations, partitions, designs, graphs), the structures it studies, and the people and machines who work with them
- Inflation (economics)
Kinds (11)
- Mathematics (the field: the practice and the structures it studies)
- Probability: weighing chances, in brains and in people's mathematics, and the structure it works on
- Intuitive statistics
- Descriptive statistics
- Statistical inference
- Bayesian statistics
- Frequentist inference
- Statistical process control
- Sequential analysis
- Census
- Machine learning
Notes published science
- Kea parrots reason like statisticians: they pick the jar with the better ratio of good tokens, not the one with more, and they discount a person who draws in a biased way. Birds and humans split at least 312 million years ago, so statistical inference may have evolved twice (src:bastos-taylor-2020).
- In 1662 John Graunt counted about 13,000 funerals a year in London's parish bills, knew three people died per eleven families a year and guessed eight per family, and got London at about 384,000: the first known ratio estimator, from a pile of death notices (src:wp:history-of-statistics).
- Same numbers, different format: when a screening test's statistics were given as probabilities, 21% of 160 gynecologists got the right answer; as natural frequencies ('8 out of 1000'), 87% did (src:hoffrage-2015).
- Stars run thermodynamics backwards: a star or star cluster that loses energy gets hotter, and two such systems touching never reach equilibrium, 'one gets hotter and hotter by losing energy, the other gets for ever colder by gaining energy'. Lynden-Bell recalls that statistical physicists first 'thought I was talking nonsense' (src:lynden-bell-1999-negative-heat-capacity).
- Erasing a single bit of memory has a minimum heat price, about 3 billionths of a trillionth of a joule at room temperature; experimenters erased bits stored in one glass bead held by lasers and watched the heat fall to that floor only when they erased slowly (src:berut-2015-landauer).
- Star clusters get hotter when they lose energy; that 'negative heat capacity' lets a cluster's core run away to ever higher temperatures (the gravothermal catastrophe), and puzzled statistical physicists who thought it was nonsense (src:lynden-bell-1999-negative-specific-heat).
- Your own eye runs on statistical mechanics: recordings of 40 retinal ganglion cells show weak correlations between pairs but strongly collective firing, and the model that fits them best is the same Ising model physicists use for magnets; pairs alone capture about 90% of the network's structure (src:schneidman-2006-retina-ising).
- The Hamiltonian's 'energy' has wandered far from joules: chemists use Gibbs free energy, and engineers have put the same form on electricity markets, macroeconomics, predator-prey ecology and the training of neural networks, where H has no energy meaning at all (src:munoz-arias-2026-port-hamiltonian).
- Bose-Einstein condensation is an overflow: once every excited state holds as many atoms as its statistics allow, 'any extra particle added will go into the ground state' (src:wp:bose-einstein-condensate).
- Economists measure growth from outer space: for the countries with the worst statistics, the best estimate of how fast incomes grow gives satellite pictures of lights at night about as much weight as the official accounts, and the answer can differ from the official figure by up to 3 percentage points a year (src:henderson-2012).
- Experimental philosophy, the young field that brought surveys into philosophy, turned out more reliable than the psychology it borrowed from: about 70% of a representative 40 studies replicated, and its papers had fewer misreported statistics (6.25% of p-values against psychology's 9.7%). Yet one of its first headline claims, that East Asian and Western students judge 'knowledge' differently, did not survive replication, and its own authors reversed course (src:cova-2021-xphi-replication, src:colombo-2018, src:wp:experimental-philosophy).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
statistics: data from a population to a decision (reader's view)
- L1: A population varying from member to member: people of a country, atoms of a crystal, a process's outputs over time, draws from a jar
- L2: Collection: census, survey sample, designed experiment, registers; in a brain, the senses taking in a sample
- L3: Data: values on nominal, ordinal, interval or ratio scales, carrying random and systematic error
- L4: Intuitive statistics in brains: statistical learning, ensemble summaries, inference from samples (the port where the discipline starts)
- L5: Description: summaries (mean, spread), tables, graphs, life tables; sufficient statistics keep all the information
- L6: Inference: models, estimators, tests, intervals, Bayesian updating; the science of uncertainty
- L7: Decision under uncertainty: policy, medicine, industry and science act on the conclusions
the boardroom (shared floors)
- L4: Intuitive statistics: statistical learning, ensemble summaries and inference from samples, in infants, great apes and kea; in a model of the hippocampus, the entorhinal-CA1 pathway
- L5: Descriptive statistics: summaries (mean, spread), tables, graphs, life tables, control charts
- L6: Inferential statistics: estimators, tests, confidence intervals, Bayesian updating, statistical models (probability as structure)
- L7: Decision under uncertainty: policy, medicine, industry, science
way in: Draws: samples seen taken from a population (balls from a box, tokens from a jar)
- L1: A population of mixed objects and a sample drawn from it
- L3: The relative frequency in each population, not the absolute count
- L4: Inference from sample to population and back (the port)
way in: Emotion: statistics anxiety
- L1: A statistics course, exam or table met
- L6: Interpretation anxiety: reading a table
way in: Hearing: streams of syllables and tones
- L1: A continuous speech stream with no pauses between words
- L3: Primitives tracked: syllables; absolute pitch in infants, relative pitch in adults
- L4: Transitional probabilities tracked (1.0 within words, 0.33 across): word boundaries found (port)
way in: Process measures: a running process watched over time
- L1: A process whose repeated measurements naturally differ
- L2: Samples taken from the process output over time
- L4: A time series with centre line and control limits
way in: Registers and censuses: every member counted and recorded
- L1: Births, deaths, people and livestock in a territory
- L2: Enumeration and recording: the census, at least every ten years
- L3: Records kept: the parish bills of mortality
- L4: Recorded values gathered for calculation
way in: Sight: groups seen at once (sizes, orientations, faces averaged)
- L1: A group of similar objects: circles of different size, oriented lines
- L4: Ensemble summary: the average of the group, formed only with attention (the port)
way in: Stated statistics: numbers read or heard (probabilities, natural frequencies)
- L1: A screening statistic stated: 1% prevalence, 80% sensitivity, 9.6% false positives
- L3: The format: probabilities or natural frequencies
- L5: Translation into frequency trees (taught)
way in: Surveys and experiments: designed samples
- L1: A population too large to count whole
- L2: A random sample or an experiment designed
- L3: Measurements taken by the same procedure before and after a manipulation
way in: The sampler: who draws, and how (biased or random)
- L1: Experimenters drawing from jars; one biased, one random
- L4: The sampler's bias folded into the prediction
way in: Thought: prior beliefs and hypotheses brought to the data
- L1: Contested origin: where the prior comes from is not settled (see Human's thought channel); booked only as what statistics takes in, the prior probability
- L4: Prior beliefs combined with sensory input (Bayesian-brain reading, contested)
way out: Choosing on the prediction (an infant's look, a kea's chosen hand)
- L1: The choice made: the hand from the better population
- L4: Prediction of the likely sample
way out: Designing the next collection: sample size, randomisation, stopping rule
- L1: The population or process sampled
- L2: Sampling continued or stopped by the rule
- L6: Planning replicates from expected effects and variability
way out: Machines: statistical models run as programs (machine learning)
- L1: Computations run
- L6: Statistical models fitted by algorithms
way out: Policy and process change: states, clinics, firms act on conclusions
- L1: Policy acting on the population counted
- L7: Decision under uncertainty
way out: Reporting: tables, graphs, papers, published estimates
- L1: Published results
- L5: Summaries and graphs
- L7: Deciding what to present