A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics

Tobias Fritz

A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1908.07021 v8 pith:2ZM2QAB3 submitted 2019-08-19 math.ST cs.LOmath.CTmath.PRstat.TH

A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics

Tobias Fritz This is my paper

classification math.ST cs.LOmath.CTmath.PRstat.TH

keywords probabilitystatisticsconditionalsufficientcategoricalfollowingindependencemarkov

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

0 comments

read the original abstract

We develop Markov categories as a framework for synthetic probability and statistics, following work of Golubtsov as well as Cho and Jacobs. This means that we treat the following concepts in purely abstract categorical terms: conditioning and disintegration; various versions of conditional independence and its standard properties; conditional products; almost surely; sufficient statistics; versions of theorems on sufficient statistics due to Fisher--Neyman, Basu, and Bahadur. Besides the conceptual clarity offered by our categorical setup, its main advantage is that it provides a uniform treatment of various types of probability theory, including discrete probability theory, measure-theoretic probability with general measurable spaces, Gaussian probability, stochastic processes of either of these kinds, and many others.

discussion (0)

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Finite Observations, Infinite Behaviour: bicategorical semantics for stateful monoidal processes
cs.LO 2026-07 accept novelty 7.5

Behaviours of stateful monoidal processes are equivalence classes of compatible finite observations in discard bicategories, yielding functorial feedback semantics and a categorified compactness theorem for closed relations.
Convex Biproducts, Stochastic Matrices and Tape Diagrams
cs.LO 2026-07 accept novelty 7.0

Convex biproducts free-generate substochastic matrix categories isomorphic to probabilistic tape diagrams, giving a complete axiomatisation of probabilistic Boolean circuits.