Causal Inference Quotes

The invalid assumption that correlation implies cause is probably among the two or three most serious and common errors of human reasoning.

If two or more instances of the phenomenon under investigation have only one circumstance in common, the circumstance in which alone all the instances agree is the cause (or effect) of the given phenomenon.

The big mystery of Big Data is causation versus correlation.

For me the problem of induction is a problem about the world: a problem of how we, as we are now (by our present scientific lights), in a world we never made, should stand better than random, or coin-tossing chances changes of coming out right when we predict by inductions. . . .

Everything has a cause.

Econometrics may be defined as the quantitative analysis of actual economic phenomena based on the concurrent development of theory and observation, related by appropriate methods of inference.

We live in a world of cause and effect.

Much of what I do in my job is think about whether relationships we see in data are causal, as opposed to just reflecting correlations. It's exactly these issues which come up in evaluating studies in public health.

I work on the boundary between economics and statistics in this field called econometrics. Part of my interest is understanding how you use statistics in productive ways to analyze dynamic economic models.

Correlation doesn’t imply causation, but it does waggle its eyebrows suggestively and gesture furtively while mouthing ‘look over there.’

Logic cannot model causal systems, and paradox is generated when time is ignored [as in logic].

Moreover, joint occurrences tend to be better recalled than instances when the effect does not occur. The proneness to remember confirming instances, but to overlook disconfirming ones, further serves to convert, in thought, coincidences into causalities.

If we possessed a thorough knowledge of all the parts of the seed of any animal (e.g. man), we could from that alone, be reasons entirely mathematical and certain, deduce the whole conformation and figure of each of its members, and, conversely if we knew several peculiarities of this conformation, we would from those deduce the nature of its seed.

Even if it had not been possible to reproduce the disease in animals and consequently to verify the hypothesis, this simple observation would have been sufficient to demonstrate the way in which the disease was propagated.

The deduction of effect from cause is often blocked by some insuperable extrinsic obstacle: the true causes may be quite unknown. Nowhere in life is this so common as in war, where the facts are seldom fully known and the underlying motives even less so.

Forces of nature act in a mysterious manner. We can but solve the mystery by deducing the unknown result from the known results of similar events.

If we knew all the laws of Nature, we should need only fact, or the description of one actual phenomenon, to infer all the particular results at that point.

It is a truism to say that a good experiment is precisely that which spares us the exertion of thinking: the better it is, the less we have to worry about its interpretation, about what it really means.

The statistician cannot excuse himself from the duty of getting his head clear on the principles of scientific inference, but equally no other thinking man can avoid a like obligation.

But to measure cause and effect... you must ensure that a simple correlation, however tempting it may be, is not mistaken for a cause. In the 1990s the stork population of Germany increased and the German at-home birth rate rose as well. Shall we credit storks for airlifting the babies?

One of the first things taught in introductory statistics textbooks is that correlation is not causation. It is also one of the first things forgotten.

Hypotheses are nets: only he who casts will catch.

In the philosophic sense, observation shows and experiment teaches.

Trying to analyze a situation without enough data was like looking at a photograph of a ball in flight and trying to gauge its direction. Is it going up, down, sideways? Is it about to collide with a baseball bat? Is it moving at all, or is something on the blind side holding it in place? A single frame didn't mean a thing. Patterns were based on data. With enough datapoints, you could predict just about anything.

Results rarely specify their causes unambiguously…. If we are forced to infer a process only from its … results, then we are usually stymied or reduced to speculation about probabilities.