Jump to ContentJump to Main Navigation
Show Summary Details
More options …

Journal of Causal Inference

Ed. by Imai, Kosuke / Pearl, Judea / Petersen, Maya Liv / Sekhon, Jasjeet / van der Laan, Mark J.

2 Issues per year

Online
ISSN
2193-3685
See all formats and pricing
More options …

Graphoids over Counterfactuals

Judea Pearl
  • Corresponding author
  • Department of Computer Science, University of California – Los Angeles, Los Angeles, CA 90095-1596, USA
  • Email
  • Other articles by this author:
  • De Gruyter OnlineGoogle Scholar
Published Online: 2014-09-12 | DOI: https://doi.org/10.1515/jci-2014-0028

Abstract

Augmenting the graphoid axioms with three additional rules enables us to handle independencies among observed as well as counterfactual variables. The augmented set of axioms facilitates the derivation of testable implications and ignorability conditions whenever modeling assumptions are articulated in the language of counterfactuals.

Keywords: conditional independence; counterfactuals; graphoids; testable implications; consistency axiom; ignorability; mediation

1 Motivation

Consider the causal Markov chain XYZ which represents the structural equations: y=f(x,u1)(1) z=g(y,u2)(2)with u1 and u2 being omitted factors such that X,u1,u2 are mutually independent.

It is well known that, regardless of the functions f and g, this model implies the conditional independence of X and Z given Y, written as XZ|Y(3)This can be readily derived from the independence of X,u1, and u2, and it also follows from the d-separation criterion, since Y blocks all paths between X and Z.

However, the causal chain can also be encoded in the language of counterfactuals by writing Yx(u)=f(x,u1)(4) Zxy(u)=g(y,u2)=Zy(u)(5)where u stands for all omitted factors (in our case u={u1,u2}) and Yx(u) stands for the value that Y would take in unit u had X been x. Accordingly, the functional and independence assumptions embedded in the chain model translate into the following counterfactual statements: Zxy=Zy(6) XYx(7) Zxy(Yx,X)(8)Equation (6) represents the missing arrow from X to Z, while eqs (7) and (8) convey the mutual independence of X,u1, and u2.1

Assume now that we are given the three counterfactual statements (6)–(8) as a specification of some uncharted model; the question arises: Are these statements testable? In other words, is there a statistical test conducted on the observed variables X, Y, and Z that could prove the model wrong? On the one hand, none of the three defining conditions (6)–(8) is testable in isolation, because each invokes a counterfactual entity. On the other hand, the fact that the chain model of eqs (1) and (2) yields the conditional independence of eq. (3) implies that the combination of all three counterfactual statements should yield a testable implication.

This paper concerns the derivation of testable conditions like eq. (3) from counterfactual sentences like eqs (6)–(8). Whereas graphical models have the benefits of inferential tools such as d-separation [1, 2, p. 335] for deriving their testable implications, counterfactual specifications must resort to the graphoid axioms,2 which, on their own, cannot reduce subscripted expressions like eqs (6)–(8) into a subscript-free expression like eq. (3). To unveil the testable implications of counterfactual specifications, the graphoid axioms must be supplemented with additional inferential machinery.

We will first prove that eq. (3) indeed follows from eqs (6) to (8) and then tackle the general question of deriving testable sentences from any given collection of counterfactual statements of the conditional independence variety. To that end, we will augment the graphoid axioms with three auxiliary inference rules, which will enable us to remove subscripts from variables and, if feasible, derive sentences in which all variables are unsubscripted, that is, testable. These auxiliary rules will rely on the composition axiom [1, p. 229] Xw=xYxw=Yw(9)which was shown to be sound and complete relative to recursive models [6, 7].3 In the special case of W={} the axiom is known as consistency rule: X=xYx=Y(10)and is discussed by Robins [9] and Pearl [10].

2 Deriving testables from non-testables

In this section we will show that eq. (3) can be derived from eqs (6) to (8) with the help of eq. (9).

We first note that substituting eq. (6) into eq. (8) yields Zy(Yx,X)(11)which is a universally quantified formula, stating that for all z, y, y′, x, x′ in the respective domains of Z, Y, and X, the following independence condition holds Zy=z(Yx=y,X=x)(12)We next note that, for the special case of x=x, eq. (12) yields Zy=z(Yx=y,X=x)or, using eq. (10) Zy=z(Y=y,X=x)forally,z,y,x(13)This can be written succinctly as Zy(Y,X)(14)Our next task is to remove the subscript from Zy. This is done in two steps. First we apply the graphoid rule of “weak union” (i.e., W(V,S)WV|S, [1, p. 11]) to obtain Zy(Y,X)ZyX|Y(15)Second, we explicate the components of eq. (15) and write Zy(X,Y)Zy=zX=x|Y=y(16)for all y, z, x, and y′. Again, for the special case of y′ = y, eq. (10) permits us to remove the subscript from Zy and write Z=zX=x|Y=yforallx,y,z(17)Finally, since the last independency holds for all x,y, and z, we can write it in succinct notation as ZX|Ywhich is subscript free and coincides with the testable implication of eq. (3).

To summarize, we have shown that the subscripts in eq. (11) can be removed in two steps. First Zy(Yx,X)Zy(Y,X)(18)and second, Zy(Y,X)ZX|Y(19)Moreover, we see that eq. (3) follows from eq. (8) alone and does not require the exogeneity assumption expressed in eq. (7).

3 Augmented graphoid axioms

In this section, we will identify three general rules that, when added to the graphoid axioms, will enable us to derive testable implications without referring back to the consistency axiom of eq. (10). The three rules are as follows:

Rule 1 V(Xw,Yxw,S)|RV(Xw,Yw,S)|R(20)

Rule 2 VR|(Xw,Yxw,S)VR|(Xw,Yw,S)(21)

Rule 3 V(Yxw,S)|(Xw,R)V(Yw,S)|(Xw,R)(22)Rules 1 and 2 state that a subscript x can be removed from Yxw whenever Yxw stands in conjunction with Xw, be it before or after the conditioning bar. In our example we had W={}. Rule 3 states that a subscript x can be removed from Yxw whenever Xw appears in the conditioning set. The symbols V,S,R in eqs (20)–(22) stand for any set of variables, observable as well as counterfactual.

The proof of these three rules follows the path that led to the derivation of eqs (18) and (19).

For mnemonic purposes we can summarize these rules using the following shorthand:

Rule 1–2 (Xw,Yxw)(Xw,Yw)(23)

Rule 3 (Yxw|Xw)(Yw|Xw)(24)

4 Deriving ignorability relations

Unveiling testable implications is only one application of the augmented graphoid axioms in Section 3. Not less important is the ability of these axioms to justify ignorability relations which a researcher may need for deriving causal effect estimands [1, 8, 11, 12].4

Consider the sentence Zx(Yz,X) which may be implied by a certain process and assume we wish to estimate the causal effect of Z on Y, P(Yz=y) from non-experimental data. For this estimation to be unbiased, the conditional ignorability ZYz|W needs to be assumed, where W is some set of observed covariates. Using Axiom (22) we can show that W=X satisfies the ignorability assumptions and, therefore, adjustment for X will yield a bias-free estimate of the causal effect P(Yz=y). This can be shown as follows: Zx(Yz,X)ZxYz|X(using the graphoid rule of “weak union”) and by Rule (22) we obtain ZxYz|XZYz|XWe therefore can write P(Yz=y)=xP(Yz=y|X=x)P(X=x)=xP(Yz=y|Z=z,X=x)P(X=x)=xP(Y=y|Z=z,X=x)P(X=x).(25)Equation (25) is none other but the standard adjustment formula for the causal effect of Z on Y, controlling for X.

The process can also be reversed; we start with a needed, yet unsubstantiated ignorability condition, and we ask whether it can be derived from more fundamental conditions which are either explicit in the model or defensible on scientific grounds. Consider, for example, an unconfounded mediation model in which treatment X is randomized and assume we seek to estimate to effect of the mediator Z on the outcome Y. (The model is depicted in Figure 1.) Operationally, we know that the ignorability condition ZYz|X would allow us to obtain the desired effect P(Yz=y) by adjusting for X, as shown in the derivation of eq. (25). However, lacking graphs for guidance, it is not clear whether this condition follows from the assumptions embedded in the model; a formal proof is therefore needed. The assumptions explicit in the model take the form

  • (a)

    Xz=X

  • (b)

    X(Zx,Yzx)

  • (c)

    ZxYzx.

(a) states that Z does not affect X, (b) represents the assumption that X is randomized, and (c) stands for the no-confounding assumption, that is, all factors affecting Z when X is held constant are independent of those affecting Y when X and Z are held constant [1, p. 232, 343]. These factors stand precisely for the “error terms” that enter the structural equations for Z and Y, respectively; hence, they have clear process-based interpretations and avail themselves to plausibility judgments.

Unconfounded mediation model implying the conditional ignorability Z⊥⊥Yz|X$$Z\, \bot\bot\!\!\! \bot \, {Y_z}\,|\,X$$
Figure 1

Unconfounded mediation model implying the conditional ignorability ZYz|X

To show that the desired ignorability condition ZYz|X follows from (a), (b) and (c), we can use Rule 3 (eq. 22) as follows. First, the standard graphoid axioms dictate X(Zx,Yzx)andZxYzxZxYzx|XNext, applying Rule 3 twice, together with X=Xz, gives ZxYzx|XZYzx|XZYzx|XzZYz|XzZYz|Xwhich yields the desired ignorability condition.

These derivations can be skipped, of course, when we have a graphical model for guidance. The adjustment formula (25) could then be written by inspection, since X satisfies the back-door condition relative to ZY. However, researchers who mistrust graphs and insist on doing the entire analysis by algebraic methods would need to use Rules 1–3 to justify the ignorability condition from assumptions (a), (b) and (c).

5 Conclusions

Rules 1–3, when added to the graphoid axioms, allow us to process conditional-independence sentences involving counterfactuals and derive both their testable implications and implications that are deemed necessary for identifying causal effects. We conjecture that Rules 1–3 are complete in the sense that all implications derivable from the graphoid axioms together with the consistency rule (18) are also derivable using the graphoid axioms together with Rules 1–3.

Augmented graphoids are by no means a substitute for causal diagrams, since the complexity of finding a derivation using graphoid axioms may be exponentially hard [13]. Diagrams, on the other hand, offer simple graphical criteria (e.g., d-separation or back-door) for deriving testable implications and effect estimands. In reasonably sized problems, these criteria can be verified by inspection, while, in large problems, they can be computed in polynomial time [14, 15]. The secret of diagrams is that they embed all the graphoid axioms in their structure and, in effect, pre-compute all their ramifications and display them in graphical patterns.

Acknowledgment

Sander Greenland and Jin Tian provided helpful comments on an early version of this note. This research was supported in parts by grants from NIH #1R01 LM009961-01, NSF #IIS-0914211 and #IIS-1018922, and ONR #N000-14-09-1-0665 and #N00014-10-1-0933.

References

  • 1.

    Pearl J. Causality: models, reasoning, and inference, 2nd ed. New York: Cambridge University Press, 2009. Google Scholar

  • 2.

    Pearl J. Probabilistic reasoning in intelligent systems. San Mateo, CA: Morgan Kaufmann, 1988. Google Scholar

  • 3.

    Dawid A. Conditional independence in statistical theory. J R Stat Soc Ser B 1979;41:1–31. Google Scholar

  • 4.

    Spohn W. Stochastic independence, causal independence, and shieldability. J Philos Logic 1980;9:73–99. CrossrefGoogle Scholar

  • 5.

    Pearl J, Paz A. GRAPHOIDS: a graph-based logic for reasoning about relevance relations. In: Boulay BD, Hogg D, Steels L, editors. Advances in artificial intelligence-II. North-Holland, 1987:357–63. Google Scholar

  • 6.

    Galles D, Pearl J. An axiomatic characterization of causal counterfactuals. Found Sci 1998;3:151–82. CrossrefGoogle Scholar

  • 7.

    Halpern J. Axiomatizing causal reasoning. In: Cooper G, Moral S, editors. Uncertainty in artificial intelligence. San Francisco, CA: Morgan Kaufmann, 1998:202–10. Also, J Artif Intell Res 2000;12:17–37. Google Scholar

  • 8.

    Holland P. Statistics and causal inference. J Am Stat Assoc 1986;81:945–60. CrossrefGoogle Scholar

  • 9.

    Robins J. A new approach to causal inference in mortality studies with a sustained exposure period – applications to control of the healthy workers survivor effect. Math Model 1986;7:1393–512. CrossrefGoogle Scholar

  • 10.

    Pearl J. On the consistency rule in causal inference: an axiom, definition, assumption, or a theorem? Epidemiology 2010;21:872–5. Web of ScienceCrossrefGoogle Scholar

  • 11.

    Rosenbaum P, Rubin D. The central role of propensity score in observational studies for causal effects. Biometrika 1983;70:41–55. CrossrefGoogle Scholar

  • 12.

    Rubin D. Estimating causal effects of treatments in randomized and nonrandomized studies. J Educ Psychol 1974;66:688–701. CrossrefGoogle Scholar

  • 13.

    Geiger D. Graphoids: a qualitative framework for probabilistic inference. PhD thesis, Department of Computer Science, University of California – Los Angeles, Los Angeles, CA, 1990. Google Scholar

  • 14.

    Shpitser I, Pearl J. Complete identification methods for the causal hierarchy. J Mach Learn Res 2008;9:1941–79. Google Scholar

  • 15.

    Tian J, Paz A, Pearl J. Finding minimal separating sets. Technical report R-254, University of California – Los Angeles, Los Angeles, CA, 1998. Google Scholar

About the article

Published Online: 2014-09-12

Published in Print: 2014-09-01


Rules for translating graphical models to counterfactual notation are given in Pearl [1, pp. 232–4], based on the structural semantics of counterfactuals. The rules represent the omitted factors affecting any variable, say Y, by the set of counterfactuals Ypa(Y), where pa(Y) stands for the parents of Y in the diagram.

The graphoid axioms are axioms of conditional independence, first formulated by Dawid [3] and Spohn [4]. Their connections to graph connectivity and to other notions of “information relevance” were established by Pearl and Paz [5] and are described in detail in Pearl [1, pp. 78–133, 2, p. 11].

The axiom of “composition” was first stated in Holland [8, p. 968]. Its completeness rests on a few technical conditions such as uniqueness and effectiveness [7].

Reliance on the assumptions of conditional ignorability [8, 11, 12], which are cognitively formidable, is one of the major weaknesses of the potential outcome framework [1, pp. 350–1]. Axioms (20)–(22) permit us to derive needed ignorability conditions from other counterfactual statements which are perhaps more transparent.


Citation Information: Journal of Causal Inference, ISSN (Online) 2193-3685, ISSN (Print) 2193-3677, DOI: https://doi.org/10.1515/jci-2014-0028.

Export Citation

©2014 by De Gruyter. Copyright Clearance Center

Comments (0)

Please log in or register to comment.
Log in