If Correlation Doesn't Imply Causation, Then What Does - DDI

7/12/13
If correlation doesnt imply causation, then what does? | DDI
DDI
About
Data-driven intelligence
Michaels main blog
Data-driven intelligence blog
Writing
If correlation doesnt imply causation, then what does?

by Michael Nielsen on January 23, 2012
Follow Michael
Subscribe to this blog Subscribe to this blog by email
It is a commonplace of scientific discussion that correlation does not imply causation. Business Week recently ran an spoof article pointing out some amusing examples of the dangers of inferring causation from correlation. For example, the article points out that Facebooks growth has been strongly correlated with the yield on Greek government bonds: (credit) Despite this strong correlation, it would not be wise to conclude that the success of Facebook has somehow caused the current (2009-2012) Greek debt crisis, nor that the Greek debt crisis has caused the adoption of Facebook! Of course, while its all very well to piously state that correlation doesnt imply causation, it does leave us with a conundrum: under what conditions, exactly, can we use experimental data to deduce a causal relationship between two or more variables? The standard scientific answer to this question is that (with some caveats) we can infer causality from a well designed randomized controlled experiment. Unfortunately, while this answer is satisfying in principle and sometimes useful in practice, its often impractical or impossible to do a randomized controlled experiment. And so were left with the question of whether there are other procedures we can use to infer causality from experimental data. And, given that we can find more general procedures for inferring causal relationships, what does causality mean, anyway, for how we reason about a system? It might seem that the answers to such fundamental questions would have been settled long ago. In fact, they turn out to be surprisingly subtle questions. Over the past few decades, a group of scientists have developed a theory of causal inference intended to address these and other related questions. This theory can be thought of as an algebra or language for reasoning about cause and effect. Many elements of the theory have been laid out in a famous book by one of the main contributors to the theory, Judea Pearl. Although the theory of causal inference is not yet fully formed, and is still undergoing development, what has already been accomplished is interesting and worth understanding. In this post I will describe one small but important part of the theory of causal inference, a causal calculus developed by Pearl. This causal calculus is a set of three simple but powerful algebraic rules which can be used to make inferences about causal relationships. In particular, Ill explain how the causal calculus can sometimes (but not always!) be used to infer causation from a set of data, even when a randomized controlled experiment is not possible. Also in the post, Ill describe some of the limits of the causal calculus, and some of my own speculations and questions. The post is a little technically detailed at points. However, the first three sections of the post are non-technical, and I hope will be of broad interest. Throughout the post Ive included occasional Problems for the author, where I describe problems Id like to solve, or things Id like to understand better. Feel free to ignore these if you find them distracting, but I hope theyll give you some sense of what I find interesting
Michael's blog on data-driven intelligence Follow Michael on Twitter
Other places on the web

GitHub Delicious Delicious research links for book project Delicious research links for last book Google Plus
Books
Reinventing Discovery: The New Era of Networked Science (Errata)
Quantum Computation and Quantum Information
Other projects
Essays TEDxWaterloo video about open science Quantum computing for the determined Polymath wiki
Recent Posts
Why Bloom filters work the way they do How to crawl a quarter billion webpages in 40 hours Using evaluation to improve our questionanswering system How to answer a question: a simple system Lisp as the Maxwells equations of software
Recent Comments
Michael Nielsen on If correlation doesnt imply causation, then what does? just_hobyst on If correlation doesnt imply
www.michaelnielsen.org/ddi/if-correlation-doesnt-imply-causation-then-what-does/
1/33
7/12/13

about the subject. Incidentally, Im sure many of these problems have already been solved by others; Im not claiming that these are all open research problems, although perhaps some are. Theyre simply things Id like to understand better. Also in the post Ive included some exercises for the reader, and some slightly harder problems for the reader. You may find it informative to work through these exercises and problems. Before diving in, one final caveat: I am not an expert on causal inference, nor on statistics. The reason I wrote this post was to help me internalize the ideas of the causal calculus. Occasionally, one finds a presentation of a technical subject which is beautifully clear and illuminating, a presentation where the author has seen right through the subject, and is able to convey that crystalized understanding to others. Thats a great aspirational goal, but I dont yet have that understanding of causal inference, and these notes dont meet that standard. Nonetheless, I hope others will find my notes useful, and that experts will speak up to correct any errors or misapprehensions on my part. Simpsons paradox Let me start by explaining two example problems to illustrate some of the difficulties we run into when making inferences about causality. The first is known as Simpsons paradox . To explain Simpsons paradox Ill use a concrete example based on the passage of the Civil Rights Act in the United States in 1964. In the US House of Representatives, 61 percent of Democrats voted for the Civil Rights Act, while a much higher percentage, 80 percent, of Republicans voted for the Act. You might think that we could conclude from this that being Republican, rather than Democrat, was an important factor in causing someone to vote for the Civil Rights Act. However, the picture changes if we include an additional factor in the analysis, namely, whether a legislator came from a Northern or Southern state. If we include that extra factor, the situation completely reverses, in both the North and the South. Heres how it breaks down: North: Democrat (94 percent), Republican (85 percent) South: Democrat (7 percent), Republican (0 percent) Yes, you read that right: in both the North and the South, a larger fraction of Democrats than Republicans voted for the Act, despite the fact that overall a larger fraction of Republicans than Democrats voted for the Act. You might wonder how this can possibly be true. Ill quickly state the raw voting numbers, so you can check that the arithmetic works out, and then Ill explain why its true. You can skip the numbers if you trust my arithmetic. North: Democrat (145/154, 94 percent), Republican (138/162, 85 percent) South: Democrat (7/94, 7 percent), Republican (0/10, 0 percent) Overall: Democrat (152/248, 61 percent), Republican (138/172, 80 percent) One way of understanding whats going on is to note that a far greater proportion of Democrat (as opposed to Republican) legislators were from the South. In fact, at the time the House had 94 Democrats, and only 10 Republicans. Because of this enormous difference, the very low fraction (7 percent) of southern Democrats voting for the Act dragged down the Democrats overall percentage much more than did the even lower fraction (0 percent) of southern Republicans who voted for the Act. (The numbers above are for the House of Congress. The numbers were different in the Senate, but the same overall phenomenon occurred. Ive taken the numbers from Wikipedias article about Simpsons paradox , and there are more details there.) If we take a naive causal point of view, this result looks like a paradox. As I said above, the overall voting pattern seems to suggest that being Republican, rather than Democrat, was an important causal factor in voting for the Civil Rights Act. Yet if we look at the individual statistics in both the North and the South, then wed come to the
Internet of Dreams - o< Good interview with an online troll: "I am lucky to have not had people who liked me for who I was when I was truly awful" [] Algorithmic Rape Jokes in the Library of Babel | Quiet Babylon On the problems caused when you create tshirts using a data mining algorithm. As Kevin Slavin has noted, this is how our culture is increasingly being created. [] Dan Pallotta: The way we think about charity is dead wrong | Video on TED.com An extremely insightful talk about the way our society thinks about not-for-profit organizations. He highlights many of the systematic ways we hobble the not-for-profit sector. [] Agnotology: The Making and Unmaking of Ignorance - Edited by Robert N. Proctor and Londa Schiebinger Linked mostly for the title, which is the useful term "agnotology" - the study of ignorance. [] How do I create a book index? Fun discussion of the minutiae of creating a book index. [] causation, then what does? just_hobyst on If correlation doesnt imply causation, then what does? jenny on If correlation doesnt imply causation, then what does? Josh on If correlation doesnt imply causation, then what does?
Linklog
Documentaries
Visual Storytelling: The Digital Video Documentary A brief skim suggests that this is a very good basic guide to making documentaries. [] The Great Ecstasy of the Woodcarver Steiner Werner Herzog Documentary of Wolfgang Steiner, one of the world's top ski-jumpers in the 1970s. The spine of the documentary is a sequence of extraordinary shots of Steiner's jumps, taken with a pair of high-speed cameras. [] Inge Druckrey: Teaching to See A reminder of the diversity of life: the passion and learning and insight people pour into calligraphy, font design, and design more generally. [] Indie Game: The Movie (2012) - IMDb Striking for the level of emotional commitment and uncertainty (and turmoil) on the part of the creators! []
2/33
7/12/13

exact opposite conclusion. To state the same result more abstractly, Simpsons paradox is the fact that the correlation between two variables can actually be reversed when additional factors are considered. So two variables which appear correlated can become anticorrelated when another factor is taken into account. You might wonder if results like those we saw in voting on the Civil Rights Act are simply an unusual fluke. But, in fact, this is not that uncommon. Wikipedias page on Simpsons paradox lists many important and similar real-world examples ranging from understanding whether there is gender-bias in university admissions to which treatment works best for kidney stones. In each case, understanding the causal relationships turns out to be much more complex than one might at first think.
Step Into Liquid (2003) Remarkable survey of the cutting edge of surfing. We see the origins of tow-rope surfing (where surfers are pulled by jet skis into waves that are too big to paddle out to), the use of hydrofoil designs that put the board a foot or two _above_ the wave, and even the use of weather stations to monitor conditions out in mid-ocean, looking for giant waves. []
Search
Ill now go through a second example of Simpsons paradox, the kidney stone treatment example just mentioned, because it helps drive home just how bad our intuitions about statistics and causality are. Imagine you suffer from kidney stones, and your Doctor offers you two choices: treatment A or treatment B. Your Doctor tells you that the two treatments have been tested in a trial, and treatment A was effective for a higher percentage of patients than treatment B. If youre like most people, at this point youd say Well, okay, Ill go with treatment A. Heres the gotcha. Keep in mind that this really happened. Suppose you divide patients in the trial up into those with large kidney stones, and those with small kidney stones. Then even though treatment A was effective for a higher overall percentage of patients than treatment B, treatment B was effective for a higher percentage of patients in both groups , i.e., for both large and small kidney stones. So your Doctor could just as honestly have said Well, you have large [or small] kidney stones, and treatment B worked for a higher percentage of patients with large [or small] kidney stones than treatment A. If your Doctor had made either one of these statements, then if youre like most people youd have decided to go with treatment B, i.e., the exact opposite treatment. The kidney stone example relies, of course, on the same kind of arithmetic as in the Civil Rights Act voting, and its worth stopping to figure out for yourself how the claims I made above could possibly be true. If youre having trouble, you can click through to the Wikipedia page , which has all the details of the numbers. Now, Ill confess that before learning about Simpsons paradox, I would have unhesitatingly done just as I suggested a naive person would. Indeed, even though Ive now spent quite a bit of time pondering Simpsons paradox, Im not entirely sure I wouldnt still sometimes make the same kind of mistake. I find it more than a little mind-bending that my heuristics about how to behave on the basis of statistical evidence are obviously not just a little wrong, but utterly, horribly wrong. Perhaps Im alone in having terrible intuition about how to interpret statistics. But frankly I wouldnt be surprised if most people share my confusion. I often wonder how many people with real decision-making power politicians, judges, and so on are making decisions based on statistical studies, and yet they dont understand even basic things like Simpsons paradox. Or, to put it another way, they have not the first clue about statistics. Partial evidence may be worse than no evidence if it leads to an illusion of knowledge, and so to overconfidence and certainty where none is justified. Its better to know that you dont know. Correlation, causation, smoking, and lung cancer As a second example of the difficulties in establishing causality, consider the relationship between cigarette smoking and lung cancer. In 1964 the United States Surgeon General issued a report claiming that cigarette smoking causes lung cancer. Unfortunately, according to Pearl the evidence in the report was based primarily on correlations between cigarette smoking and lung cancer. As a result the report came under attack not just by tobacco companies, but also by some of the Search
Archives
September 2012 August 2012 June 2012 April 2012 March 2012 February 2012 January 2012 July 2011 June 2011
3/33
7/12/13

worlds most prominent statisticians, including the great Ronald Fisher. They claimed that there could be a hidden factor maybe some kind of genetic factor which caused both lung cancer and people to want to smoke (i.e., nicotine craving). If that was true, then while smoking and lung cancer would be correlated, the decision to smoke or not smoke would have no impact on whether you got lung cancer. Now, you might scoff at this notion. But derision isnt a principled argument. And, as the example of Simpsons paradox showed, determining causality on the basis of correlations is tricky, at best, and can potentially lead to contradictory conclusions. Itd be much better to have a principled way of using data to conclude that the relationship between smoking and lung cancer is not just a correlation, but rather that there truly is a causal relationship. One way of demonstrating this kind of causal connection is to do a randomized, controlled experiment. We suppose there is some experimenter who has the power to intervene with a person, literally forcing them to either smoke (or not) according to the whim of the experimenter. The experimenter takes a large group of people, and randomly divides them into two halves. One half are forced to smoke, while the other half are forced not to smoke. By doing this the experimenter can break the relationship between smoking and any hidden factor causing both smoking and lung cancer. By comparing the cancer rates in the group who were forced to smoke to those who were forced not to smoke, it would then be possible determine whether or not there is truly a causal connection between smoking and lung cancer. This kind of randomized, controlled experiment is highly desirable when it can be done, but experimenters often dont have this power. In the case of smoking, this kind of experiment would probably be illegal today, and, I suspect, even decades into the past. And even when its legal, in many cases it would be impractical, as in the case of the Civil Rights Act, and for many other important political, legal, medical, and econonomic questions. Causal models To help address problems like the two example problems just discussed, Pearl introduced a causal calculus. In the remainder of this post, I will explain the rules of the causal calculus, and use them to analyse the smoking-cancer connection. Well see that even without doing a randomized controlled experiment its possible (with the aid of some reasonable assumptions) to infer what the outcome of a randomized controlled experiment would have been, using only relatively easily accessible experimental data, data that doesnt require experimental intervention to force people to smoke or not, but which can be obtained from purely observational studies. To state the rules of the causal calculus, well need several background ideas. Ill explain those ideas over the next three sections of this post. The ideas are causal models (covered in this section), causal conditional probabilities , and d-separation, respectively. Its a lot to swallow, but the ideas are powerful, and worth taking the time to understand. With these notions under our belts, well able to understand the rules of the causal calculus To understand causal models, consider the following graph of possible causal relationships between smoking, lung cancer, and some unknown hidden factor (say, a hidden genetic factor):
4/33
7/12/13
This is a quite general model of causal relationships, in the sense that it includes both the suggestion of the US Surgeon General (smoking causes cancer) and also the suggestion of the tobacco companies (a hidden factor causes both smoking and cancer). Indeed, it also allows a third possibility: that perhaps both smoking and some hidden factor contribute to lung cancer. This combined relationship could potentially be quite complex: it could be, for example, that smoking alone actually reduces the chance of lung cancer, but the hidden factor increases the chance of lung cancer so much that someone who smokes would, on average, see an increased probability of lung cancer. This sounds unlikely, but later well see some toy model data which has exactly this property. Of course, the model depicted in the graph above is not the most general possible model of causal relationships in this system; its easy to imagine much more complex causal models. But at the very least this is an interesting causal model, since it encompasses both the US Surgeon General and the tobacco company suggestions. Ill return later to the possibility of more general causal models, but for now well simply keep this model in mind as a concrete example of a causal model. Mathematically speaking, what do the arrows of causality in the diagram above mean? Well develop an answer to that question over the next few paragraphs. It helps to start by moving away from the specific smoking-cancer model to allow a causal model to be based on a more general graph indicating possible causal relationships between a number of variables:
Each vertex in this causal model has an associated random variable, For example, in the causal model above someone smokes or gets lung cancer, and and
could be a two-outcome random variable indicates smokes or does not smoke,
indicating the presence or absence of some gene that exerts an influence on whether indicates gets lung cancer or doesnt get lung cancer. The other variables would refer to other potential dependencies in this (somewhat more
complex) model of the smoking-cancer connection. A notational convention that well use often is to interchangeably use to refer to a
random variable in the causal model, and also as a way of labelling the corresponding vertex in the graph for the causal model. It should be clear from context which is meant. Well also sometimes refer interchangeably to the causal model or to the associated graph. For the notion of causality to make sense we need to constrain the class of graphs that can be used in a causal model. Obviously, itd make no sense to have loops in
5/33
7/12/13
the graph:
We cant have
causing
causing
causing
! At least, not without a time
machine. Because of this we constrain the graph to be a directed acyclic graph, meaning a (directed) graph which has no loops in it. By the way, I must admit that Im not a fan of the term directed acyclic graph. It sounds like a very complicated notion, at least to my ear, when what it means is very simple: a graph with no loops. Id really prefer to call it a loop-free graph, or something like that. Unfortunately, the directed acyclic graph nomenclature is pretty standard, so well go with it. Our picture so far is that a causal model consists of a directed acyclic graph, whose vertices are labelled by random variables variables. Intuitively, what causality means is that for any particular variables which directly influence the value of collection the only random , i.e., the . For . To complete our definition of causal models we need to capture the allowed relationships between those random
are the parents of
of random variables which are connected directly to :
instance, in the graph shown below (which is the same as the complex graph we saw a little earlier), we have
Now, of course, vertices further back in the graph say, the parents of the parents could, of course, influence the value of mediated through the parent vertices. Note, by the way, that Ive overloaded the notation, using (or , or to denote a ) to denote a . But it would be indirect, an influence
collection of random variables. Ill use this kind of overloading quite a bit in the rest of this post. In particular, Ill often use the notation subset of random variables from the graph. Motivated by the above discussion, one way we could define causal influence would be to require that be a function of its parents:
where
is some function. In fact, well allow a slightly more general notion of to not just be a deterministic function of the parents, but be expressible in the form:
causal influence, allowing
a random function. We do this by requiring that
where
is a function, and
is a collection of random variables such that: (a) the is . , except when (and, through is itself, or a descendant of
are independent of one another for different values of ; and (b) for each , independent of all variables The intuition is that the some extra randomness into
are a collection of auxiliary random variables which inject , its descendants), but which are
6/33
7/12/13

otherwise independent of the variables in the causal model. Summing up, a causal model consists of a directed acyclic graph, are labelled by random variables, another, and each descendant of . or the auxiliary random , and each . The for some function , whose vertices
is expressible in the form are independent of one , except when is or a
is independent of all variables
In practice, we will not work directly with the functions variables
. Instead, well work with the following equation, which specifies the
causal models joint probability distribution as a product of conditional probabilities:
I wont prove this equation, but the expression should be plausible, and is pretty easy to prove; Ive asked you to prove it as an optional exercise below. Exercises Prove the above equation for the joint probability distribution. Problems (Simpsons paradox in causal models) Consider the causal model of smoking introduced above. Suppose that the hidden factor is a gene which is either switched on or off. If on, it tends to make people both smoke and get lung cancer. Find explicit values for conditional probabilities in the causal model such that is, we have both and . Problems for the author An alternate, equivalent approach to defining causal models is as follows: (1) all root vertices (i.e., vertices with no parents) in the graph are labelled by independent random variables. (2) augment the graph by introducing new vertices corresponding to the pointing to . These new vertices have single outgoing edges, . (3) Require that non-root vertices in the augmented graph be , and yet if the additional genetic factor is taken into account this relationship is reversed. That
deterministic functions of their parents. The disadvantage of this definition is that it introduces the overhead of dealing with the augmented graph. But the definition also has the advantage of cleanly separating the stochastic and deterministic components, and I wouldnt be surprised if developing the theory of causal inference from this point of view was stimulating, at the very least, and may possibly have some advantages compared to the standard approach. So the problem I set myself (and anyone else who is interested!) is to carry the consequences of this change through the rest of the theory of causal inference, looking for advantages and disadvantages. Ive been using terms like causal influence somewhat indiscriminately in the discussion above, and so Id like to pause to discuss a bit more carefully about what is meant here, and what nomenclature we should use going forward. All the arrows in a causal model indicate are the possibility of a direct causal influence. This results in two caveats on how we think about causality in these models. First, it may be that a child random variable is actually completely independent of the value of one (or more) of its parent random variables. This is, admittedly, a rather special case, but is perfectly consistent with the definition. For example, in a causal model like
7/33
7/12/13
it is possible that the outcome of cancer might be independent of the hidden causal factor or, for that matter, that it might be independent of whether someone smokes or not. (Indeed, logically, at least, it may be independent of both, although of course thats not what well find in the real world.) The second caveat in how we think about the arrows and causality is that the arrows only capture the direct causal influences in the model. It is possible that in a causal model like
will have a causal influence on
through its influence on
and
. This would
be an indirect causal influence, mediated by other random variables, but it would still be a causal influence. In the next section Ill give a more formal definition of causal influence that can be used to make these ideas precise. Causal conditional probabilities In this section Ill explain what I think is the most imaginative leap underlying the causal calculus. Its the introduction of the concept of causal conditional probabilities . The notion of ordinary conditional probabilities is no doubt familiar to you. Its pretty straightforward to do experiments to estimate conditional probabilities such as , simply by looking at the population of people who smoke, and figuring out what fraction of those people develop cancer. Unfortunately, for the purpose of understanding the causal relationship between smoking and cancer, isnt the quantity we want. As the tobacco companies pointed out, there might well be a hidden genetic factor that makes it very likely that youll see cancer in anyone who smokes, but that wouldnt therefore mean that smoking causes cancer. As we discussed earlier, what youd really like to do in this circumstance is a randomized controlled experiment in which its possible for the experimenter to force someone to smoke (or not smoke), breaking the causal connection between the hidden factor and smoking. In such an experiment you really could see if there was a causal influence by looking at what fraction of people who smoked got cancer. In particular, if that fraction was higher than in the overall population then youd be justified in concluding that smoking helped cause cancer. In practice, its probably not practical to do this kind of randomized controlled experiment. But Pearl had what turns out to be a very clever idea: to imagine a hypothetical world in which it really is possible to force someone to (for example) smoke, or not smoke. In particular, he introduced a conditional causal probability , which is the conditional probability of cancer in this hypothetical world. This should be read as the (causal conditional) probability of cancer given that we do smoking, i.e., someone
8/33
7/12/13

has been forced to smoke in a (hypothetical) randomized experiment. Now, at first sight this appears a rather useless thing to do. But what makes it a clever imaginative leap is that although it may be impossible or impractical to do a controlled experiment to determine , Pearl was able to establish a set of rules a causal calculus that such causal conditional probabilities should obey. And, by making use of this causal calculus, it turns out to sometimes be possible to infer the value of probabilities such as even when a controlled, randomized experiment is impossible. And thats a very remarkable thing to be able to do, and why I say it was so clever to have introduced the notion of causal conditional probabilities. Well discuss the rules of the causal calculus later in this post. For now, though, lets develop the notion of causal conditional probabilities. Suppose we have a causal model of some phenomenon: ,
Now suppose we introduce an external experimenter who is able to intervene to deliberately set the value of a particular variable equivalent to having a new causal model: to . In other words, the experimenter can override the other causal influences on that variable. This is
In this new causal model, weve represented the experimenter by a new vertex, which has as a child the vertex the parents to to of . All other parents of are cut off, i.e., the edges from are deleted from the graph. In this case that means the edge from
has been deleted. This represents the fact that the experimenters are left undisturbed.) In fact, its even simpler (and equivalent) to consider a , and no extra vertex added:
intervention overrides the other causal influences. (Note that the edges to the children causal model where the parents have been cut off from
9/33
7/12/13
This model has no vertex explicitly representing the experimenter, but rather the relation this graph by is replaced by the relation . We will denote have been , indicating the graph in which all edges pointing to
deleted. We will call this a perturbed graph, and the corresponding causal model a perturbed causal model. In the perturbed causal model the only change is to delete the edges to . Our aim is to use this perturbed causal model to compute the conditional causal probability the causal conditional probability the probability distribution in the perturbed causal model, probability distribution in the original causal model was given by . In this expression, indicates that the is just the value of . To term is omitted before the , since the value of is set on the right. By definition, , and to replace the relation by the relation
compute the value of the probability in the perturbed causal model, note that the
where the product on the right is over all vertices in the causal model. This expression remains true for the perturbed causal model, but a single term on the right-hand side changes: the conditional probability for the changed from result we have: term. In particular, this term gets to be . As a to , since we have fixed the value of
This equation is a fundamental expression, capturing what it means for an experimenter to intervene to set the value of some particular variable in a causal model. It can easily be generalized to a situation where we partition the variables into two sets, and , where are the variables we suppose have been set by are intervention in a (possibly hypothetical) randomized controlled experiment, and the remaining variables:
Note that on the right-hand side the values for the appropriate values from
are assumed to be given by
and . The expression [1] can be viewed as a definition
of causal conditional probabilities. But although this expression is fundamental to understanding the causal calculus, it is not always useful in practice. The problem is that the values of some of the variables on the right-hand side may not be known, and cannot be determined by experiment. Consider, for example, the case of smoking and cancer. Recall our causal model:
10/33
7/12/13
What wed like is to compute
. Unfortunately, we
immediately run into a problem if we try to use the expression on the right of equation [1]: weve got no way of estimating the conditional probabilities for smoking given the hidden common factor. So we cant obviously compute whenever were worried about the possible influence of some hidden factor. All is not lost, however. Just because we cant compute the expression on the right of [1] directly doesnt mean we cant compute causal conditional probabilities in other ways, and well see below how the causal calculus can help solve this kind of problem. Its not a complete solution we shall see that it doesnt always make it possible to compute causal conditional probabilities. But it does help. In particular, well see that although its not possible to compute this causal model, it is possible to compute similar causal model, one that still has a hidden factor. With causal conditional probabilities defined, were now in position to define more precisely what we mean by causal influence. Suppose we have a causal model, and and we say are distinct random variables (or disjoint subsets of random variables). Then has a causal influence over if there are values and of and of . In other words, an external experimenter can cause a corresponding change in the . The following exercise gives an information-theoretic can transmit information to if and only if the above condition for in a very . And, as you can perhaps imagine, this is the kind of problem that will come up a lot
such that who can intervene to change the value of distribution of values at can intervene to set
justification for this definition of causal influence: it shows that an experimenter who for causal inference is met. Exercises (The causal capacity) This exercise is for people with some background in information theory. Suppose we define the causal capacity between be indicate that the value of , where is the mutual information, the for (we use the hat to is the maximization is over possible distributions corresponding random variable at and to
is being set by intervention), and , with distribution
. Shannons noisy channel coding theorem tells us that an external experimenter who can intervene to set the value of transmit information to an observer at a causal influence over . capacity. Show that the causal capacity is greater than zero if and only if can has at a maximal rate set by the causal
Weve just defined a notion of causal influence between two random variables in a causal model. What about when we say something like Event A causes Event B? What does this mean? Returning to the smoking-cancer example, it seems that we would say that cancer causes smoking provided , so that if someone makes the choice to smoke, uninfluenced by other causal factors, then they would increase their chance of cancer. Intuitively, it seems to me that this notion of events causing one another should be related to the notion of causal influence just defined above. But I dont yet see quite how to do that. The first problem below suggests a conjecture in this direction:
11/33
7/12/13
Problems for the author Suppose and
are random variables in a causal model such that for some pair of values and . Does exerts a causal influence on ?
this imply that
(Sum-over-paths for causal conditional probabilities?) I believe a kind of sum-over-paths formulation of causal conditional probabilities is possible, but havent worked out details. The idea is as follows (the details may be quite wrong, but I believe something along these lines should work). Supose would like to show first that if ancestor of paths from then to in is not an ancestor of then is an and are single vertices (with corresponding random variables) in a causal model. Then I , i.e., intervention does nothing. Second, if
may be obtained by summing over all directed , and computing for each path a contribution to the
sum which is a product of conditional probabilities along the path. (Note that we may need to consider the same path multiple times in the sum, since the random variables along the path may take different values). We used causal models in our definition of causal conditional probabilities. But our informal definiton imagine a hypothetical world in which its possible to force a variable to take a particular value didnt obviously require the use of a causal model. Indeed, in a real-world randomized controlled experiment it may be that there is no underlying causal model. This leads me to wonder if there is some other way of formalizing the informal definition weve given? Another way of framing the last problem is that Im concerned about the empirical basis for causal models. How should we go about constructing such models? Are they fundamental, representing true facts about the world, or are they modelling conveniences? (This is by no means a dichotomy.) It would be useful to work through many more examples, considering carefully the origin of the functions and of the auxiliary random variables d-separation In this section well develop a criterion that Pearl calls directional separation (dseparation, for short). What d-separation does is let us inspect the graph of a causal model and conclude that a random variable about the value of another random variable in the model cant tell us anything in the model, or vice versa. .
To understand d-separation well start with a simple case, and then work through increasingly complex cases, building up our intuition. Ill conclude by giving a precise definition of d-separation, and by explaining how d-separation relates to the concept of conditional independence of random variables. Heres the first simple causal model:
Clearly, knowing
can in general tell us something about and
in this kind of causal and
model, and so in this case are d-connected.
are not d-separated. Well use the term d-
connected as a synonym for not d-separated, and so in this causal model
By contrast, in the following causal model
and
dont give us any information
about each other, and so they are d-separated:
A useful piece of terminology is to say that a vertex like the middle vertex in this model is a collider for the path from to , meaning a vertex at which both edges
12/33
7/12/13
along the path are incoming. What about the causal model:
In this case, it is possible that knowing
will tell us something about
, because of
their common ancestry. Its like the way knowing the genome for one sibling can give us information about the genome of another sibling, since similarities between the genomes can be inferred from the common ancestry. Well call a vertex like the middle vertex in this model a fork for the path from which both edges are outgoing. Exercises Construct an explicit causal model demonstrating the assertion of the last paragraph. For example, you may construct a causal model in which are joined by a fork, and where Suppose we have a path from colliders along the path, and let that is actually a function of to . be the number of and to , meaning a vertex at
in a causal model. Let or
be the number of forks along the path. Show , i.e., the number of forks and
can only take the values
colliders is either the same or differs by at most one. Well say that a path (of any length) from to that contains a collider is a block ed
path. By contrast, a path that contains no colliders is called an unblock ed path. (Note that by the above exercise, an unblocked path must contain either one or no forks.) In general, we define and to be d-connected if there is an unblocked path between them. We define them to be d-separated if there is no such unblocked path. Its worth noting that the concepts of d-separation and d-connectedness depend only on the graph topology and on which vertices and have been chosen. In and , merely particular, they dont depend on the nature of the random variables
on the identity of the corresponding vertices. As a result, you can determine dseparation or d-connectdness simply by inspecting the graph. This fact that dseparation and d-connectdness are determined by the graph also holds for the more sophisticated notions of d-separation and d-connectedness we develop below. With that said, it probably wont surprise you to learn that the concept of d-separation is closely related to whether or not the random variables exercises. Ill state a much more general connection below. Exercises Suppose that and are d-separated. Show that . . Explain how and and are independent and are independent of one another. This is a connection you can (optionally) develop through the following
random variables, i.e., that
Suppose we have two vertices which are d-connected in a graph corresponding to those two vertices are not independent.
to construct a causal model on that graph such that the random variables
The last two exercises almost but dont quite claim that random variables
and
in a causal model are independent if and only if they are d-separated. Why does this statement fail to be true? How can you modify the statement to make it true? So far, this is pretty simple stuff. It gets more complicated, however, when we extend the notion of d-separation to cases where we are conditioning on already k nowing the value of one or more random variables in the causal model. Consider, for example, the graph:
13/33
7/12/13
(Figure A.) Now, if we know , then knowing doesnt give us any additional information about is already a function of and . So it makes sense to and are d-
, since by our original definition of a causal model say that blocks this path from . to
some auxiliary random variables which are independent of
, even though in the unconditioned case this
path would not have been considered blocked. Well also say that separated, given
It is helpful to give a name to vertices like the middle vertex in Figure A, i.e., to vertices with one ingoing and one outgoing edge. Well call such vertices a traverse along the path from is that if to . Using this language, the lesson of the above discussion to , then the path is blocked. is in a traverse along a path from
By contrast, consider this model:
In this case, knowing we know . unblocked path from
will in general give us additional information about blocks one path from and to to . And so we say that
, even if
. This is because while
there is another
are d-connected, given
Another case similar to Figure A is the model with a fork:
Again, if we know about to
, then knowing
as well doesnt give us any extra information is blocking the path from are d-separated, given to . ,
(or vice versa). So well say that in this case and
, even though in the unconditioned case this path would not have been
considered blocked. Again, in this example The lesson of this model is that if then the path is blocked. A subtlety arises when we consider a collider:
is located at a fork along a path from
(Figure B.) In the unconditioned case this would have been considered a blocked path. And, naively, it seems as though this should still be the case: at first sight (at least according to my intuition) it doesnt seem very likely that information about (or vice versa), even given that can give us any additional is known. Yet we should be and some auxiliary
cautious, because the argument we made for the graph in Figure A breaks down: we cant say, as we did for Figure A, that independent random variables. In fact, were wise to be cautious because and really can tell us something is a function of
extra about one another, given a knowledge of
. This is a phenomenon which Pearl
calls Berk sons paradox. He gives the example of a graduate school in music which
14/33
7/12/13
undergraduate grades (encoded in

will admit a student (a possibility encoded in the value of exceptionally gifted at music (encoded in ) if either they have high ) or some other evidence that they are ). It would not be surprising if these two
attributes were anticorrelated amongst students in the program, e.g., students who were admitted on the basis of exceptional gifts would be more likely than otherwise to have low grades. And so in this case knowledge of knowledge of accepted into the program). Another way of seeing Berksons paradox is to construct an explicit causal model for the graph in Figure B. Consider, for example, a causal model in which independent random bits, that of , where or , chosen with equal probabilities and are . We suppose , knowing the value (exceptional gifts) would give us (they were (likely to have low grades), conditioned on knowledge of
is addition modulo . This causal model does, indeed, , since .
have the structure of Figure B. But given that we know the value tells us everything about
As a result of this discussion, in the causal graph of Figure B well say that unblocks the path from and to , even though in the unconditioned case the path . and can tell us and where the would have been considered blocked. And well also say that in this causal graph are d-connected, conditional on
The immediate lesson from the graph of Figure B is that something about one another, given only collider is at
, if there is a path between
. In fact, the same phenomenon can occur even in this graph:
(Figure C.) To see this, suppose we choose i.e., independent random bits, the unlabelled vertex be see as before that . The general intuition about graphs like that in Figure C is that knowing infer something about the ancestors of from to , since ancestors are known, too. As a result, in this case we say that And so in this case is d-connected to , given . allows us to and as in the example just described above, . We will let . Then we , because
or , chosen with equal probabilities . And, finally, we choose , given that we know
can tell us something about
, and so we must act as though those unblocks the path to .
has an ancestor which is a collider on the path from
Given the discussion of Figure C that weve just had, you might wonder why forks or traverses which are ancestors of instance, why dont we consider graph: cant block a path, for similar reasons? For and to be d-separated, given , in the following
The reason, of course, is that its easy to construct examples where something about in addition to what we already know from
tells us
. And so we cant
15/33
7/12/13
consider and

to be d-separated, given , in this example.
These examples motivate the following definition: Definition: Let , and be disjoint subsets of vertices in a causal model. to a vertex in . We say the path is block ed by , or (b) a fork
Consider a path from a vertex in which is in
if the path contains either: (a) a collider which is not an ancestor of , or (c) a traverse which is in and blocked. We say that given are d-connected, given and some vertex in .
. We say the path is unblock ed if it is not , if there is an unblocked and are d-separated,
path between some vertex in
, if they are not d-connected. and are d-separated given is a bit of a mouthful, and so its helpful to . Note that this
Saying
have an abbreviated notation. Well use the abbreviation notation includes the graph clear. Well write to denote unconditional d-separation.
; well sometimes omit the graph when the context is
As an aside, Pearl uses a similar but slightly different notation for d-separation, namely . Unfortunately, while the symbol looks like a LaTeX symbol, its not, but is most easily produced using a rather dodgy LaTeX hack. Instead of using that hack over and over again, Ive adopted a more standard LaTeX notation. While Im making asides, let me make a second: when I was first learning this material, I found the d for directional in d-separation and d-connected rather confusing. It suggested to me that the key thing was having a directed path from one vertex to the other, and that the complexities of colliders, forks, and so on were a sideshow. Of course, theyre not, theyre central to the whole discussion. For this reason, when I was writing these notes I considered changing the terminology to iseparated and i-connected, for informationally-separated and informationallyconnected. Ultimately I decided not to do this, but I thought mentioning the issue might be helpful, in part to reassure readers (like me) who thought the d seemed a little mysterious. Okay, thats enough asides, lets get back to the main track of discussion. We saw earlier that (unconditional) d-separation is closely connected to the independence of random variables. It probably wont surprise you to learn that conditional d-separation is closely connected to conditional independence of random variables. Recall that two sets of random variables independent, given a third set of random variables independence occurs in a causal model: Theorem (graphical criterion for conditional independence): Let and let , and be disjoint subsets of vertices in that graph. Then , if and only if for all causal models on and are conditionally independent, given . d-separated, given corresponding to be a graph, and are and , if are conditionally .
The following theorem shows that d-separation gives a criterion for when conditional
the random variables
(Update: Thank s to Rob Spek k ens for pointing out an error in my original statement of this theorem.) I wont prove the theorem here. However, its not especially difficult if youve followed the discussion above, and is a good problem to work through: Problems Prove the above theorem. Problems for the author The concept of d-separation plays a central role in the causal calculus. My sense is that it should be possible to find a cleaner and more intuitive definition that substantially simplifies many proofs. Itd be good to spend some time trying to
16/33
7/12/13
find such a definition. The causal calculus
Weve now got all the concepts we need to state the rules of the causal calculus. There are three rules. The rules look complicated at first, although theyre easy to use once you get familiar with them. For this reason Ill start by explaining the intuition behind the first rule, and how you should think about that rule. Having understood how to think about the first rule its easy to get the hang of all three rules, and so after that Ill just outright state all three rules. In what follows, we have a causal model on a graph perturbed graph in which all edges pointing to value of , overriding other causal influences on , and are disjoint denotes the have been
subsets of the variables in the causal model. Recall also that
from the parents of .
deleted. This is the graph which results when an experimenter intervenes to set the
Rule 1: When can we ignore observations: Ill begin by stating the first rule in all its glory, but dont worry if you dont immediately grok the whole rule. Instead, just take a look, and try to start getting your head around it. What well do then is look at some simple special cases, which are easily understood, and gradually build up to an understanding of what the full rule is saying. Okay, so heres the first rule of the causal calculus. What it tells us is that when , then we can ignore the observation of probability of , conditional on both in computing the : and an intervention to set
To understand why this rule is true, and what it means, lets start with a much simpler case. Lets look at what happens to the rule when there are no from in the original (unperturbed) graph , so and or variables in is d-separated the mix. In this case, our starting assumption simply becomes that because theres no circumstance we have the rule in this case is merely that standard definition of being independent. and
. Theres no need to worry about is independent of . But the statement of
variable whose value is being set by intervention. In this , which is, indeed, equivalent to the
In other words, the first rule is simply a generalization of what it means for (1) by adding in an extra variable
to be independent. The full rule generalizes the notion of independence in two ways: whose value has been determined by passive whose value has been set by observation; and (2) by adding in an extra variable paragraphs. We begin with generalization (1), i.e., there is no our starting assumption becomes that graph section this means that variable in the mix. In this case, , given , in the , and so
intervention. Well consider these two ways of generalizing separately in the next two
is d-separated from
. By the graphical criterion for conditional independence discussed in the last is conditionally independent of , given and , which is exactly the statement of the rule. And so the first rule
can be viewed as a generalization of what it means for conditional on .
to be independent,
Now lets look at the other generalization, (2), in which weve added an extra variable whose value has been set by intervention, and where there is no mix. In this case, our starting assumption becomes that given value of , in the perturbed graph conditional indepenence tells us that variable in the , is d-separated from
. In this case, the graphical criterion for is independent from , conditional on the
being set by experimental intervention, and so . Again, this is exactly the statement of the rule.
The full rule, of course, merely combines both these generalizations in the obvious way. It is really just an explicit statement of the content of the graphical criterion for conditional independence, in a context where has been observed, and the value of
17/33
7/12/13

set by experimental intervention. The rules of the causal calculus: All three rules of the causal calculus follow a similar template to the first rule: they provide ways of using facts about the causal structure (notably, d-separation) to make inferences about conditional causal probabilities. Ill now state all three rules. The intuition behind rules 2 and 3 wont necessarily be entirely obvious, but after our discussion of rule 1 the remaining rules should at least appear plausible and comprehensible. Ill have bit more to say about intuition below. As above, we have a causal model on a graph subsets of the variables in the causal model. which all edges pointing to from the parents of the graph in which all edges pointing out from deleted. We will also freely use notations like these operations. Rule 1: When can we ignore observations: Suppose . Then: , and are disjoint denotes
denotes the perturbed graph in have been deleted. to the children of have been
to denote combinations of
Rule 2: When can we ignore the act of intervention: Suppose . Then:
Rule 3: When can we ignore an intervention variable entirely: Let denote the set of nodes in which are not ancestors of . Suppose . Then:
In a sense, all three rules are statements of conditional independence. The first rule tells us when we can ignore an observation. The second rule tells us when we can ignore the act of intervention (although that doesnt necessarily mean we can ignore the value of the variable being intervened with). And the third rule tells us when we can ignore an intervention entirely, both the act of intervention, and the value of the variable being intervened with. I wont prove rule 2 or rule 3 this post is already quite long enough. (If I ever significantly revise the post I may include the proofs). The important thing to take away from these rules is that they give us conditions on the structure of causal models so that we know when we can ignore observations, acts of intervention, or even entire variables that have been intervened with. This is obviously a powerful set of tools to be working with in manipulating conditional causal probabilities! Indeed, according to Pearl theres even a sense in which this set of rules is complete, meaning that using these rules you can identify all causal effects in a causal model. I havent yet understood the proof of this result, or even exactly what it means, but thought Id mention it. The proof is in papers by Shpitser and Pearl and Huang and Valtorta . If youd like to see the proofs of the rules of the calculus, you can either have a go at proving them yourself, or you can read the proof. Problems for the author Suppose the conditions of rules 1 and 2 hold. Can we deduce that the conditions of rule 3 also hold? Using the causal calculus to analyse the smoking-lung cancer connection Well now use the causal calculus to analyse the connection between smoking and lung cancer. Earlier, I introduced a simple causal model of this connection:
18/33
7/12/13
The great benefit of this model was that it included as special cases both the hypothesis that smoking causes cancer and the hypothesis that some hidden causal factor was responsible for both smoking and cancer. It turns out, unfortunately, that the causal calculus doesnt help us analyse this model. Ill explain why thats the case below. However, rather than worrying about this, at this stage its more instructive to work through an example showing how the causal calculus can be helpful in analysing a similar but slightly modified causal model. So although this modification looks a little mysterious at first, for now I hope youll be willing to accept it as given. The way Im going to modify the causal model is by introducing an extra variable, namely, whether someone has appreciable amounts of tar in their lungs or not:
(By tar, I dont mean tar literally, but rather all the material deposits found as a result of smoking.) This causal model is a plausible modification of the original causal model. It is at least plausible to suppose that smoking causes tar in the lungs and that those deposits in turn cause cancer. But if the hidden causal factor is genetic, as the tobacco companies argued was the case, then it seems highly unlikely that the genetic factor caused tar in the lungs, except by the indirect route of causing those people to smoke. (Ill come back to what happens if you refuse to accept this line of reasoning. For now, just go with it.) Our goal in this modified causal model is to compute probabilities like . What well show is that the causal calculus lets us compute this probability entirely in terms of probabilities like and other probabilities that dont involve an intervention, i.e., that dont involve . without needing to
This means that we can determine hidden factor. It also means that we can determine
know anything about the hidden factor. We wont even need to know the nature of the without needing to intervene to force someone to smoke or not smoke, i.e., to set the value for .
In other words, the causal calculus lets us do something that seems almost miraculous: we can figure out the probability that someone would get cancer given that they are in the smoking group in a randomized controlled experiment, without needing to do the randomized controlled experiment. And this is true even though there may be a hidden causal factor underlying both smoking and cancer. Okay, so how do we compute ?
19/33
7/12/13
the conditional causal probability

The obvious first question to ask is whether we can apply rule 2 or rule 3 directly to .
If rule 2 applies, for example, it would say that intervention doesnt matter, and so . Intuitively, this seems unlikely. Wed expect that intervention really can change the probability of cancer given smoking, because intervention would override the hidden causal factor. If rule 3 applies, it would say that more unlikely than rule 2 applying. However, as practice and a warm up, lets work through the details of seeing whether rule 2 or rule 3 can be applied directly to For rule 2 to apply we need . . To check whether this is true, recall that deleted: , i.e., that an intervention to
force someone to smoke has no impact on whether they get cancer. This seems even
is the graph with the edges pointing out from
Obviously, influence both
is not d-separated from and
in this graph, since
and
have a
common ancestor. This reflects the fact that the hidden causal factor indeed does . So we cant apply rule 2. . Recall that is the
What about rule 3? For this to apply wed need graph with the edges pointing toward deleted:
Again,
is not d-separated from to
, in this case because we have an unblocked path can influence ,
directly from
. This reflects our intuition that the value of
even when the value of
has been set by intervention. So we cant apply rule 3. .
Okay, so we cant apply the rules of the causal calculus directly to determine Is there some indirect way we can determine this probability? An experienced probabilist would at this point instinctively wonder whether it would help to condition on the value of , writing:
Of course, saying an experienced probabilist would instinctively do this isnt quite the same as explaining why one should do this! However, it is at least a moderately obvious thing to do: the only extra information we potentially have in the problem is , and so its certainly somewhat natural to try to introduce that variable into the problem. As we shall see, this turns out to be a wise thing to do. Exercises I used without proof the equation . This should be intuitively plausible, but really requires proof. Prove that the equation is correct. To simplify the right-hand side of equation [2], we first note that we can apply rule 2 to the second term on the right-hand side, obtaining this explicitly, note that the condition for rule 2 to apply is that . To check . We
20/33
7/12/13
already saw the graph graph, since the only path from

above, and, indeed, to is d-separated from in that is blocked at . As a result, we have:
At this point in the presentation, Im going to speed the discussion up, telling you what rule of the calculus to apply at each step, but not going through the process of explicitly checking that the conditions of the rule hold. (If youre doing a close read, you may wish to check the conditions, however.) The next thing we do is to apply rule 2 to the first term on the right-hand side of equation [3], obtaining to remove the gives us: , obtaining . We then apply rule 3 . Substituting back in
So this means that weve reduced the computation of of
to the computation
. This doesnt seem terribly encouraging: weve merely substituted the
computation of one causal conditional probability for another. Still, let us continue plugging away, and see if we can make progress. The obvious first thing to try is to apply rule 2 or rule 3 to simplify . Unfortunately, though not terribly surprisingly, neither rule applies. So what do we do? Well, in a repeat of our strategy above, we again condition on the other variable we have available to us, in this case :
Now were cooking! Rule 2 lets us simplify the first term to us simplify the second term to , and so we have
, while rule 3 lets
. To substitute this expression back into equation [4] it helps to change the summation index from have a duplicate summation index. This gives us: to , since otherwise we would
This is the promised expression for
(i.e., for probabilities like
, assuming the causal model above) in terms of quantities which may be observed directly from experimental data, and which dont require intervention to do a randomized, controlled experiment. Once is determined, we can compare it against is larger than smoking does, indeed, play a causal role in cancer. Something that bugs me about the derivation of equation [5] is that I dont really know how to see through the calculations. Yes, it all works out in the end, and its easy enough to follow along. Yet thats not the same as having a deep understanding. Too many basic questions remain unanswered: Why did we have to condition as we did in the calculation? Was there some other way we could have proceeded? What would have happeed if wed conditioned on the value of the hidden variable? (This is not obviously the wrong thing to do: maybe the hidden variable would ultimately drop out of the calculation). Why is it possible to compute causal probabilities in this model, but not (as we shall see) in the model without tar? Ideally, a deeper understanding would make the answers to some or all of these questions much more obvious. Problems for the author Why is it so much easier to compute than in the model then we can conclude that . If
above? Is there some way we could have seen that this would be the case, without needing to go through a detailed computation? Suppose we have a causal model of for which subsets and , with a subset of vertices for which all
conditional probabilities are known. Is it possible to give a simple characterization of vertices it is possible to compute ? using just the conditional probabilities from
Unfortunately, I dont know what the experimentally observed probabilities are in the
21/33
7/12/13

smoking-tar-cancer case. If anyone does, Id be interested to know. In lieu of actual data, Ill use some toy model data suggested by Pearl; the data is quite unrealistic, but nonetheless interesting as an illustration of the use of equation [5]. The toy model data is as follows: (1) 47.5 percent of the population are nonsmokers with no tar in their lungs, and 10 percent of these get cancer. (2) 2.5 percent are smokers with no tar, and 90 percent get cancer. (3) 2.5 percent are nonsmokers with tar, and 5 percent get cancer. (4) 47.5 percent are smokers with tar, and 85 percent get cancer. In this case, we get:
By contrast,
percent, and so if this data was correct (obviously
its not even close) it would show that smoking actually somewhat reduces a persons chance of getting lung cancer. This is despite the fact that percent, and so a naive approach to causality based on correlations alone would suggest that smoking causes cancer. In fact, in this imagined world smoking might actually be useable as a preventative treatment for cancer! Obviously this isnt truly the case, but it does illustrate the power of this method of analysis. Summing up the general lesson of the smoking-cancer example, suppose we have two competing hypotheses for the causal origin of some effect in a system, A causes C or B causes C, say. Then we should try to construct a realistic causal model which includes both hypotheses, and then use the causal calculus to attempt to distinguish the relative influence of the two causal factors, on the basis of experimentally accessible data. Incidentally, the kind of analysis of smoking we did above obviously wasnt done back in the 1960s. I dont actually know how causality was established over the protestations that correlation doesnt impy causation. But its not difficult to think of ways you might have come up with truly convincing evidence that smoking was a causal factor. One way would have been to look at the incidence of lung cancer in populations where smoking had only recently been introduced. Suppose, for example, that cigarettes had just been introduced into the (fictional) country of Nicotinia, and that this had been quickly followed by a rapid increase in rates of lung cancer. If this pattern was seen across many new markets then it would be very difficult to argue that lung cancer was being caused solely by some pre-existing factor in the population. Exercises Construct toy model data where smoking increases a persons chance of getting lung cancer. Lets leave this model of smoking and lung cancer, and come back to our original model of smoking and lung cancer:
What would have happened if wed tried to use the causal calculus to analyse this model? I wont go through all the details, but you can easily check that whatever rule
22/33
7/12/13

you try to apply you quickly run into a dead end. And so the causal calculus doesnt seem to be any help in analysing this problem. This example illustrates some of the limitations of the causal calculus. In order to compute particular structure: we needed to assume a causal model with a
While this model is plausible, it is not beyond reproach. You could, for example, criticise it by saying that it is not the presence of tar deposits in the lungs that causes cancer, but maybe some other factor, perhaps something that is currently unknown. This might lead us to consider a causal model with a revised structure:
So we could try instead to use the causal calculus to analyse this new model. I havent gone through this exercise, but I strongly suspect that doing so we wouldnt be able to use the rules of the causal calculus to compute the relevant probabilities. The intuition behind this suspicion is that we can imagine a world in which the tar may be a spurious side-effect of smoking that is in fact entirely unrelated to lung cancer. What causes lung cancer is really an entirely different mechanism, but we couldnt distinguish the two from the statistics alone. The point of this isnt to say that the causal calculus is useless. Its remarkable that we can plausibly get information about the outcome of a randomized controlled experiment without actually doing anything like that experiment. But there are limitations. To get that information we needed to make some presumptions about the causal structure in the system. Those presumptions are plausible, but not logically inevitable. If someone questions the presumptions then it may be necessary to revise the model, perhaps adopting a more sophisticated causal model. One can then use the causal calculus to attempt to analyse that more sophisticated model, but we are not guaranteed success. It would be interesting to understand systematically when this will be possible and when it will not be. The following problems start to get at some of the issues involved. Problems for the author Is it possible to make a more precise statement than the causal calculus doesnt seem to be any help for the original smoking-cancer model? Given a probability distribution over some random variables, it would be useful to have a classification theorem describing all the causal models in which those
23/33
7/12/13

random variables could appear. Extending the last problem, itd be good to have an algorithm to answer questions like: in the space of all possible causal models consistent with a given set of observed probabilities, what can we say about the possible causal probabilities? It would also be useful to be able to input to the algorithm some constraints on the causal models, representing knowledge were already sure of. In real-world experiments there are many practical issues that must be addressed to design a realiable randomized, controlled experiment. These issues include selection bias, blinding, and many others. There is an entire field of experimental design devoted to addressing such issues. By comparison, my description of causal inference ignores many of these practical issues. Can we integrate the best thinking on experimental design with ideas such as causal conditional probabilities and the causal calculus? From a pedagogical point of view, I wonder if it might have been better to work fully through the smoking-cancer example before getting to the abstract statement of the rules of the causal calculus. Those rules can all be explained and motivated quite nicely in the context of the smoking-cancer example, and that may help in understanding. Conclusion Ive described just a tiny fraction of the work on causality that is now going on. My impression as an admittedly non-expert outsider to the field is that this is an exceptionally fertile field which is developing rapidly and giving rise to many fascinating applications. Over the next few decades I expect the theory of causality will mature, and be integrated into the foundations of disciplines ranging from economics to medicine to social policy. Causal discovery: One question Id like to understand better is how to discover causal structures inside existing data sets. After all, human beings do a pretty good (though far from perfect) job at figuring out causal models from their observation of the world. Id like to better understand how to use computers to automatically discover such causal models. I understand that there is already quite a literature on the automated discovery of causal models, but I havent yet looked in much depth at that literature. I may come back to it in a future post. Im particularly fascinated by the idea of extracting causal models from very large unstructured data sets. The KnowItAll group at the University of Washington (see Oren Etzioni on Google Plus) have done fascinating work on a related but (probably) easier problem, the problem of open information extraction. This means taking an unstructured information source (like the web), and using it to extract facts about the real world. For instance, using the web one would like computers to be able to learn facts like Barack Obama is President of the United States, without needing a human to feed it that information. One of the things that makes this task challenging is all the misleading and difficult-to-understand information out on the web. For instance, there are also webpages saying George Bush is President of the United States, which was probably true at the time the pages were written, but which is now misleading. We can find webpages which state things like [Let's imagine] Steve Jobs is President of the United States; its a difficult task for an unsupervised algorithm to figure out how to interpret that Lets imagine. What the KnowItAll team have done is made progress on figuring out how to learn facts in such a rich but uncontrolled environment. What Im wondering is whether such techniques can be adapted to extract causal models from data? Itd be fascinating if so, because of course humans dont just reason with facts, they also reason with (informal) causal models that relate those facts. Perhaps causal models or a similar concept may be a good way of representing some crucial part of our knowledge of the world.
24/33
7/12/13
Problems for the author
What systematic causal fallacies do human beings suffer from? We certainly often make mistakes in the causal models we extract from our observations of the world one example is that we often do assume that correlation implies causation, even when thats not true and itd be nice to understand what systematic biases we have. Humans arent just good with facts and causal models. Were also really good at juggling multiple causal models, testing them against one another, finding problems and inconsistencies, and making adjustments and integrating the results of those models, even when the results conflict. In essence, we have a (working, imperfect) theory of how to deal with causal models. Can we teach machines to do this kind of integration of causal models? We know that in our world the sun rising causes the rooster to crow, but its possible to imagine a world in which it is the rooster crowing that causes the sun to rise. This could be achieved in a suitably designed virtual world, for example. The reason we believe the first model is correct in our world is not intrinsic to the data we have on roosters and sunrise, but rather depends on a much more complex network of background knowledge. For instance, given what we know about roosters and the sun we can easily come up with plausible causal mechanisms (solar photons impinging on the roosters eye, say) by which the sun could cause the rooster to crow. There do not seem to be any similarly plausible causal models in the other direction. How do we determine what makes a particular causal model plausible or not? How do we determine the class of plausible causal models for a given phenomenon? Can we make this kind of judgement automatically? (This is all closely related to the last problem). Continuous-time causality: A peculiarity in my post is that even though were talking about causality, and time is presumably important, Ive avoided any explicit mention of time. Of course, its implicitly there: if Id been a little more precise in specifying my models theyd no doubt be conditioned on events like smoked at least a pack a day for 10 or more years. Of course, this way of putting time into the picture is rather coarse-grained. In a lot of practical situations were interested in understanding causality in a much more temporally fine-grained way. To explain what I mean, consider a simple model of the relationship between what we eat and our insulin levels:
This model represents the fact that what we eat determines our insulin levels, and our insulin levels in turn play a part in determining how hungry we feel, and thus what we eat. But as a model, its quite inadequate. In fact, theres a much more complex feedback relationship going on, a constant back-and-forth between what we eat at any given time, and our insulin levels. Ideally, this wouldnt be represented by a few discrete events, but rather by a causal model that reflects the continual feedback
25/33
7/12/13

between these possibilities. What Id like to see developed is a theory of continuoustime causal models, which can address this sort of issue. It would also be useful to extend the calculus to continuous spaces of events. So far as I know, at present the causal calculus doesnt work with these kinds of ideas. Problems for the author Can we formulate theories like electromagnetism, general relativity and quantum mechanics within the framework of the causal calculus (or some generalization)? Do we learn anything by doing so? Other notions of causality: A point Ive glossed over in the post is how the notion of causal influence weve been studying relates to other notions of causality. The notion weve been exploring is based on the notion of causality that is established by a (hopefully well-designed!) randomized controlled experiment. To understand what that means, think of what it would mean if we used such an experiment to establish that smoking does, indeed, cause cancer. All this means is that in the population being studied, forcing someone to smoke will increase their chance of getting cancer. Now, for the practical matter of setting public health policy, thats obviously a pretty important notion of causality. But nothing says that we wont tomorrow discover some population of people where no such causal influence is found. Or perhaps well find a population where smoking actively helps prevent cancer. Both these are entirely possible. Whats going on is that while our notion of causality is useful for some purposes, it doesnt necessarily say anything about the details of an underlying causal mechanism, and it doesnt tell us how the results will apply to other populations. In other words, while its a useful and important notion of causality, its not the only way of thinking about causality. Something Id like to do is to understand better what other notions of causality are useful, and how the intervention-based approach weve been exploring relates to those other approaches. Acknowledgments Thanks to Jen Dodd, Rob Dodd, and Rob Spekkens for many discussions about causality. Especial thanks to Rob Spekkens for pointing me toward the epilogue of Pearls book, which is what got me hooked on causality! Principal sources and further reading A readable and stimulating overview of causal inference is the epilogue to Judea Pearls book. The epilogue, in turn, is based on a survey lecture by Pearl on causal inference. I highly recommend getting a hold of the book and reading the epilogue; if you cannot do that, I suggest looking over the survey lecture. A draft copy of the first edition of the entire book is available on Pearls website. Unfortunately, the draft does not include the full text of the epilogue, only the survey lecture. The lecture is still good, though, so you should look at it if you dont have access to the full text of the epilogue. Ive also been told good things about the book on causality by Spirtes, Glymour and Scheines, but havent yet had a chance to have a close look at it. An unfortunate aspect of the current post is that it gives the impression that the theory of causal inference is entirely Judea Pearls creation. Of course thats far from the case, a fact which is quite evident from both Pearls book, and the Spirtes-GlymourScheines book. However, the particular facets Ive chosen to focus on are due principally to Pearl and his collaborators: most of the current post is based on chapter 3 and chapter 1 of Pearls book, as well as a 1994 paper by Pearl, which established many of the key ideas of the causal calculus. Finally, for an enjoyable and informative discussion of some of the challenges involved in understanding causal inference I recommend Jonah Lehrers recent article in Wired. Interested in more? Please follow me on Twitter. You may also enjoy reading my
26/33

If Correlation Doesn't Imply Causation, Then What Does - DDI

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

If Correlation Doesn't Imply Causation, Then What Does - DDI

Uploaded by

Copyright:

Available Formats

7/12/13

If correlation doesnt imply causation, then what does? | DDI

Michaels main blog

Data-driven intelligence blog

If correlation doesnt imply causation, then what does?

Michael's blog on data-driven intelligence Follow Michael on Twitter

Other places on the web

Reinventing Discovery: The New Era of Networked Science (Errata)

Quantum Computation and Quantum Information

If correlation doesnt imply causation, then what does? | DDI

If correlation doesnt imply causation, then what does? | DDI

If correlation doesnt imply causation, then what does? | DDI

If correlation doesnt imply causation, then what does? | DDI

could be a two-outcome random variable indicates smokes or does not smoke,

If correlation doesnt imply causation, then what does? | DDI

! At least, not without a time

are the parents of

of random variables which are connected directly to :

causal influence, allowing

a random function. We do this by requiring that

If correlation doesnt imply causation, then what does? | DDI

is expressible in the form are independent of one , except when is or a

is independent of all variables

In practice, we will not work directly with the functions variables

causal models joint probability distribution as a product of conditional probabilities:

If correlation doesnt imply causation, then what does? | DDI

will have a causal influence on

through its influence on

If correlation doesnt imply causation, then what does? | DDI

If correlation doesnt imply causation, then what does? | DDI

are assumed to be given by

and . The expression [1] can be viewed as a definition

If correlation doesnt imply causation, then what does? | DDI

What wed like is to compute

is being set by intervention), and , with distribution

If correlation doesnt imply causation, then what does? | DDI

this imply that

can in general tell us something about and

in this kind of causal and

model, and so in this case are d-connected.

are not d-separated. Well use the term d-

connected as a synonym for not d-separated, and so in this causal model

By contrast, in the following causal model

dont give us any information

about each other, and so they are d-separated:

If correlation doesnt imply causation, then what does? | DDI

In this case, it is possible that knowing

will tell us something about

in a causal model. Let or

can only take the values

random variables, i.e., that

If correlation doesnt imply causation, then what does? | DDI

some auxiliary random variables which are independent of

, even though in the unconditioned case this

By contrast, consider this model:

In this case, knowing we know . unblocked path from

. This is because while

are d-connected, given

Another case similar to Figure A is the model with a fork:

Again, if we know about to

(or vice versa). So well say that in this case and

is located at a fork along a path from