Home > Uncategorized > Getting causality into statistics

## Getting causality into statistics

Lars Syll

Because statistical analyses need a causal skeleton to connect to the world, causality is not extra-statistical but instead is a logical antecedent of real-world inferences. Claims of random or “ignorable” or “unbiased” sampling or allocation are justified by causal actions to block (“control”) unwanted causal effects on the sample patterns. Without such actions of causal blocking, independence can only be treated as a subjective exchangeability assumption whose justification requires detailed contextual information about absence of factors capable of causally influencing both selection (including selection for treatment) and outcomes. Otherwise it is essential to consider pathways for the causation of biases (nonrandom, systematic errors) and their interactions …

Probability is inadequate as a foundation for applied statistics, because competent statistical practice integrates logic, context, and probability into scientific inference and decision, using causal narratives to explain diverse data. Thus, given the absence of elaborated causality discussions in statistics textbooks and coursework, we should not be surprised at the widespread misuse and misinterpretation of statistical methods and results. This is why incorporation of causality into introductory statistics is needed as urgently as other far more modest yet equally resisted reforms involving shifts in labels and interpretations for P-values and interval estimates.

Sander Greenland

Causality can never be reduced to a question of statistics or probabilities unless you are — miraculously — able to keep constant all other factors that influence the probability of the outcome studied. To understand causality we always have to relate it to a specific causal structure. Statistical correlations are never enough. No structure, no causality.

Statistical patterns should never be seen as anything else than possible clues to follow. Behind observable data, there are real structures and mechanisms operating, things that are  — if we really want to understand, explain and (possibly) predict things in the real world — more important to get hold of than to simply correlate and regress observable variables.

Statistics cannot establish the truth value of a fact. Never has. Never will.

1. March 8, 2023 at 9:42 pm

Getting causality into statistics
The morning after a bank crisis, what’s dangerous to bank systems is always what’s perceived as risky.
Before the crisis, the excessive bank exposures causing it, are always built-up with what’s perceived as very safe.
http://subprimeregulations.blogspot.com/2019/03/my-letter-to-financial-stability-board.html

2. March 11, 2023 at 9:06 am

I agree with everything in this post by Lars except the last sentence, which does not follow from the prior argument. A causal structure should have empirical implications. Statistics can never prove the hypothesized structure is “true”. But if its implications can be shown not to hold with a very high degree of confidence by statistical testing the causal structure can be rejected. The statement “this causal structure is inconsistent with the evidence and is false or incomplete” is one whose truth value can therefore be established.
Popper’s distinction applies for all practical purposes: a universal statement can be disproved but not proved; an existential statement can be proved but nor disproved. Statistics, like everything else, is subject to that maxim.

• March 12, 2023 at 8:05 am

《The statement “this causal structure is inconsistent with the evidence and is false or incomplete” is one whose truth value can therefore be established.》

Didn’t epicyclists use this principle to disprove Aristarchus’s 3rd century BC heliocentric solar system theory, because parallax could not be observed? (Sure, Aristarchus’s theory was incomplete and wildly underestimated distances, but were epicycles even more wrong, despite passing the parallax test which was used to disprove heliocentricism?

Consider Feyerabend (from Wikipedia):

《Feyerabend offers several criticisms of empiricism. One concerns the distinction between observational and theoretical terms. If an observational term is understood as one whose acceptance can be determined by immediate perception, then what counts as ‘observational’ or ‘theoretical’ changes throughout history as our patterns of habituation change and our ability to directly perceive entities evolve.[61] On another definition, observation terms are those that can be known directly and with certainty whereas theoretical terms are hypothetical. Feyerabend argues that all statements are hypothetical, since the act of observation requires theories to justify its veridicality.》

Does Feyerabend’s call for scientific pluralism remind anyone else of Lars’s call for economic pluralism?

3. March 21, 2023 at 11:53 am

Feyerarbend makes points at the margin, like Quine’s “two dogmas of empiricism”. For 99 per cent of practical purposes Popper’s distinction holds and that you cannot derive an ought from an is remains a serviceable principle too. Empiricsm does not guarantee perfection but, as Clement Atlee said when asked what it felt like to be 80 years old, it’s: “better than the alternative”. Aristarchus was not alone. There have been many correct speculations that could not displace an accepted theory because the technology of testing or observation was not sufficiently advanced at the time.
I am strongly in favour of economic pluralism and share most of Lars’ views of contemporary economics. I differ from his defeatism about current methods of empirical testing. The problem is not that there are no empirical findings but that orthodoxy too often ignores them when inconvenient. Raising arcane philosophical points about the basis of mathematics or statistics is fun but can aid the obscurantists who want to ignore evidence.,

• March 25, 2023 at 8:03 am

“For 99 per cent of practical purposes Popper’s distinction holds”

Could you have said that when epicycles ruled, or when Aristotelian physics allowed bridges that still stand to be built, or when geologists denied that continents move? How do you know when the idea you’ve rejected is part of that 1%?

Is it just some sort of Expected Value calculation? Are you figuring that it is smarter to bet on the consensus, because the emotional consequences of ostracism are so dire to you?

“Clement Atlee said when asked what it felt like to be 80 years old, it’s: “better than the alternative””

What does this have to do with my brother (suicide at 49) or father (suicide at 72), who voluntarily preferred the alternative? Again, are your arguments in favor of the consensus just based on emotions, not any kind of real truth?

“There have been many correct speculations that could not displace an accepted theory because the technology of testing or observation was not sufficiently advanced at the time.”

But why ban me from a geological forum, for example, for questioning that discipline’s assumptions, unless there is really some emotional issue of controlling the narrative at work, separate from any supposedly rational, emotionless scientific quest for Truth?

“orthodoxy too often ignores them when inconvenient”

Don’t heterodox economists such as Lars post studies that support their theories, just when it’s convenient?

“aid the obscurantists who want to ignore evidence.”

What has the consensus done for me, and why is the evidence I provide for my viewpoint regularly scrubbed from the record (present company thankfully excepted, so far)?