pith. sign in

arxiv: 1907.01672 · v1 · pith:QXJJYL26new · submitted 2019-07-02 · 🧮 math.ST · stat.TH

Causal models on probability spaces

Irineo Cabreros , John D. Storey This is my paper

Pith reviewed 2026-05-25 10:07 UTC · model grok-4.3

classification 🧮 math.ST stat.TH
keywords causal inferencepotential outcomesmeasure theoryprobability spacescausal modelsrandomizationmatching
0
0 comments X

The pith

Causal models built on probability spaces clarify effects, interactions, matching and randomization.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper constructs causal models in the potential outcomes framework directly on probability spaces using the tools of measure theory. This construction is meant to demonstrate that measure theory supplies a precise language for expressing and reasoning about causality. A sympathetic reader would care because the approach is shown to yield concrete insights into familiar causal concepts through simple examples. The authors also introduce a visualization method and take initial steps toward an axiomatic theory of general causal models.

Core claim

By constructing causal models on probability spaces within the potential outcomes framework, measure theory provides a precise and instructive language for causality, and consideration of the probability spaces underlying causal models offers clarity into central concepts of causal inference. Simple examples demonstrate insights into causal effects, causal interactions, matching procedures, and randomization. A visualization technique is introduced that aids both example generation and causal intuition, and an axiomatic framework is supplied as initial steps toward a formal theory of general causal models.

What carries the argument

The construction of causal models on probability spaces, in which potential outcomes are represented as random variables on a measure space so that causal quantities become measurable functions and expectations.

If this is right

  • Causal effects are expressed as expectations of the difference between potential-outcome random variables over the probability space.
  • Causal interactions appear as properties of the joint distribution of multiple potential-outcome variables.
  • Matching procedures correspond to measurable partitions or conditioning operations on the sample space.
  • Randomization corresponds to independence between the treatment-assignment variable and the potential-outcome variables.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same construction might let researchers import limit theorems or concentration inequalities directly into causal bounds.
  • Extending the representation to infinite or non-separable spaces could handle continuous or uncountably many interventions without new machinery.
  • The axiomatic sketch could serve as a common language for comparing potential-outcomes models with structural causal models.

Load-bearing premise

Every causal model of interest in the potential outcomes framework can be faithfully represented as a standard probability space without requiring additional structure or losing essential causal features.

What would settle it

A specific causal model in the potential outcomes framework whose counterfactual relations or inference properties cannot be preserved when embedded into any probability space.

Figures

Figures reproduced from arXiv: 1907.01672 by Irineo Cabreros, John D. Storey.

Figure 1
Figure 1. Figure 1: A binary random variable X on the square sample. Shaded regions denote the pre-image of 1. Black points correspond to individual elements of the sample space. Ω x y R2 0 1 0 1 (X, Y )(ω) [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A system of two binary random variables X and Y defined on the square space. The pre-image of 1 for the random variable X is the upper half of Ω. The pre-image of 1 for the random variable Y is the upper right triangle. then X and Y are called independent. Otherwise, X and Y are dependent. In [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The probability space (Ω, F, P) is formed by the product of spaces the spaces (Ω1, F1, P1) and (Ω2, F2, P2). The two binary random variables X and Y defined separately on the original probability spaces are independent on the product space. are independent random variables when defined jointly on the product space (Ω1 × Ω2, F1 × F2, P1 × P2) [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (a) The system of random variables X and Y from [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Experimental randomization of X induces a product space structure. 4 Causal inference on three variables Several new concepts in causality arise in systems of three observable variables. As such, in this section we study the simplest three-variable system: three binary random variables. We add to our running example of smoking (X) and lung cancer (Y ) a third binary random variable Z representing exercise … view at source ↗
Figure 6
Figure 6. Figure 6: A system of three random variables X, Y , and Z for which X and Z are jointly causal for Y , but neither is individually causal for Y . However, the double-subscripted potential outcomes {Y(X,Z)=(x,z)} differ from each other on a subset of Ω of measure one. This is because for all ω ∈ Ω (excluding the measure zero subset along the vertical and horizontal mid-line of Ω), exactly one double-subscripted poten… view at source ↗
Figure 7
Figure 7. Figure 7: A set of potential outcomes {Y(X,Z)=(x,z)}(x,z) for which Z is causal for YX=0 and YX=1, however X and Z are not jointly causal for Y . follows: Definition 4 (Joint causality). Two binary random variables X and Z are said to be jointly causal for a third random variable Y if both of the following hold: (i) Z is causal for YX=x for some x. (ii) X is causal for YZ=z for some z. According to Definition 4, X a… view at source ↗
Figure 8
Figure 8. Figure 8: (left) The observable random variables X and Y as in [PITH_FULL_IMAGE:figures/full_fig_p020_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visual representation of Axioms 1 and 2. The complete pot [PITH_FULL_IMAGE:figures/full_fig_p024_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Visual representation of Axiom 3. The partial potential [PITH_FULL_IMAGE:figures/full_fig_p025_10.png] view at source ↗
read the original abstract

We describe the interface between measure theoretic probability and causal inference by constructing causal models on probability spaces within the potential outcomes framework. We find that measure theory provides a precise and instructive language for causality and that consideration of the probability spaces underlying causal models offers clarity into central concepts of causal inference. By closely studying simple, instructive examples, we demonstrate insights into causal effects, causal interactions, matching procedures, and randomization. Additionally, we introduce a simple technique for visualizing causal models on probability spaces that is useful both for generating examples and developing causal intuition. Finally, we provide an axiomatic framework for causality and make initial steps towards a formal theory of general causal models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper constructs causal models on probability spaces within the potential outcomes framework, arguing that measure theory supplies a precise and instructive language for causality while examination of the underlying probability spaces clarifies central concepts such as causal effects, interactions, matching procedures, and randomization. It supports this through simple examples, a visualization technique for generating examples and developing intuition, and an axiomatic framework that takes initial steps toward a formal theory of general causal models.

Significance. If the constructions faithfully embed counterfactuals, interventions, and identifiability without loss of essential features, the work could strengthen foundational understanding in causal inference by grounding it more explicitly in measure theory. Credit is due for the concrete examples illustrating insights into matching and randomization, the visualization technique as a practical tool, and the explicit axiomatic framework as a step toward formalization; these elements provide pedagogical and conceptual value even if the framework remains preliminary.

minor comments (3)
  1. [Abstract] The abstract states that the constructions 'offer clarity into central concepts' but does not indicate the specific section or proposition where this clarity is demonstrated beyond the examples; adding a forward reference would improve readability.
  2. Notation for the probability spaces (e.g., the definition of the sample space, sigma-algebra, and measure in the causal model constructions) should be cross-referenced to standard measure-theory texts to aid readers unfamiliar with the interface.
  3. The visualization technique is described as useful for generating examples, but the manuscript would benefit from an explicit algorithm or pseudocode in the relevant section to make the method reproducible.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our manuscript, the recognition of its pedagogical value through examples and visualization, and the recommendation for minor revision. No specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The paper constructs causal models directly on probability spaces by embedding the potential outcomes framework into measure-theoretic language, supplying explicit examples of causal effects, interactions, matching, and randomization along with a visualization technique and axiomatic framework. None of these steps reduces a claimed result to a fitted parameter, a self-definitional loop, or a load-bearing self-citation whose content is itself unverified; the constructions are presented as independent applications of standard measure theory to existing causal concepts. The central claim that measure theory supplies a precise language for causality therefore rests on the supplied constructions rather than on any internal reduction to the paper's own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The work rests on the standard axioms of measure-theoretic probability and the potential outcomes framework as background; no free parameters, invented entities, or ad-hoc axioms are introduced in the abstract.

axioms (2)
  • standard math Standard axioms of measure-theoretic probability spaces
    The paper builds causal models inside probability spaces, invoking the usual sigma-algebra and measure properties.
  • domain assumption Potential outcomes framework as given
    The construction is performed within the existing potential outcomes framework without re-deriving it.

pith-pipeline@v0.9.0 · 5625 in / 1247 out tokens · 24271 ms · 2026-05-25T10:07:02.919813+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [1]

    Outline of a new principle of mathematic al psychology (1851)

    Gustav Theodor Fechner. Outline of a new principle of mathematic al psychology (1851). Psychological Research, 49(4):203–207, Dec 1987

    work page 1987

  2. [2]

    Causal analysis after haave lmo

    James J Heckman and Rodrigo Pinto. Causal analysis after haave lmo. Working Paper 19453, National Bureau of Economic Research, September 2013

    work page 2013

  3. [3]

    The Grammar of Science

    Karl Pearson. The Grammar of Science . Adam & Charles Black, 3 edition, 1911

    work page 1911

  4. [4]

    Correlations genuine and spurious in pearson and yu le

    John Aldrich. Correlations genuine and spurious in pearson and yu le. Statist. Sci. , 10(4):364–376, 11 1995

    work page 1995

  5. [5]

    The Book of Why: The New Science of Cause and Effect

    Judea Pearl and Dana Mackenzie. The Book of Why: The New Science of Cause and Effect. Basic Books, 2018

    work page 2018

  6. [6]

    Marshall

    A. Marshall. Principles of Economics . Macmillan and Company, 1890

    work page

  7. [7]

    Retrospectives: Ceteris paribus

    Joseph Persky. Retrospectives: Ceteris paribus. Journal of Economic Perspectives , 4(2):187–193, June 1990

    work page 1990

  8. [8]

    J. Neyman. On the application of probability theory to agricultura l experiments. essay on principles. Statistical Science, 5(4), 1923

    work page 1923

  9. [9]

    D. B. Rubin. Estimating causal effects of treatments in randomiz ed and nonrandomized studies. Journal of Educational Psychology , 66(5), 1974

    work page 1974

  10. [10]

    P. W. Holland. Statistics and causal inference. Journal of the American Statistical Association, 81(396), 1986

    work page 1986

  11. [11]

    Haavelmo

    T. Haavelmo. The probability approach in econometrics. Econometrica, 12, 1944

    work page 1944

  12. [12]

    J. Pearl. Causality: Models, Reasoning, and Inference . Cambridge University Press, 2000

    work page 2000

  13. [13]

    Causation, Prediction, and Search

    Peter Spirtes, Clark Glymour, and Richard Scheines. Causation, Prediction, and Search. A Bradford Book, 2 edition, 2001

    work page 2001

  14. [14]

    Lauritzen

    Steffen L. Lauritzen. Causal inference from graphical models , 2001

    work page 2001

  15. [15]

    D. B. Rubin and G. W. Imbens. Causal Inference for Statistics, Social, and Biomedical Sciences: an Introduction . Cambridge University Press, 2015

    work page 2015

  16. [16]

    Rosenbaum

    Paul R. Rosenbaum. A characterization of optimal designs for observational studies. Journal of the Royal Statistical Society. Series B (Methodo logical), 53, 01 1991

    work page 1991

  17. [17]

    W. G. Cochran and S. Paul Chambers. The planning of observat ional studies of human populations. Journal of the Royal Statistical Society , 128(2):234–277, 1965

    work page 1965

  18. [18]

    treatment

    David Williams. Probability with Martingales . Cambridge University Press, 1991. 28 A Review of classical probability theory A.1 The probability space Definition 8 (probability space). A probability space, denoted (Ω, F , P ), consists of three objects: (i) Ω: A set called the sample space . (ii) F : A set of subsets of Ω. F must contain Ω and be closed un...

    work page 1991

  19. [19]

    For S = (1 , 2,

    IxS ∩ I ˜xS = ∅ for all xS ̸= ˜xS Proof. For S = (1 , 2, . . . , n) (i.e., when we are considering the statement applied to the complete potential outcomes) both 1 and 2 follow simply from the pro perties of indicator random variables. When S ⊂ { 1, 2, . . . , n}, we note that since IxS = ∑ x ¯S Ix, we have that: IxS = ∪ x ¯S Ix Part 1 is now clear becaus...

    work page