pet peeve - causal language

[gusl]

I am often annoyed by people saying things that imply a causal relationship when the latter is completely unjustified. I can recall 3 instances of this in the last month-and-a-half, one of which came from a distinguished statistician, who used the subtly causal "a high value of X is good for you". (I'm not counting a talk by a statistician doing drug research who really should worry about confounding, and who found my comment totally surprising; I didn't see a problem with his language)

If statisticians are not immune to this, how can we expect working scientists to keep the two notions separate?

Most of us say things we don't mean when it comes to causality, especially in informal settings. This perpetuates muddled thinking, and gives scientists excuse to be imprecise. So I have started a wiki page, to help authors separate causal from associational language. Think of it as a pair of Thesaurus entries, designed by a prescriptivist to have zero overlap: http://www.optimizelife.com/wiki/Causal_language

Please submit here entries here, as I'm keeping editorial control over this.

Comments

[tedesson.livejournal.com]

Do you have examples of situations which meet your justified causal relationship criteria?

This is the sort of thing I read lesswrong for.

[gustavolacerda.livejournal.com]

Causal conclusions can be justified in many ways. Some examples:
* experimental (rather than merely observational) data
* assumptions about how much has been measured (e.g., see p.10 of this PDF for a definition of "causal sufficiency")
* reasons from the scientific theory, e.g. plausible mechanisms (though some would say that this isn't statistical inference if one already had such beliefs before seeing the data)

The first two are purely statistical, i.e. once the model is given, you don't need to know what the data is about in order to infer causality, and can be performed by a computer that is blind to scientific common sense. (There is something to be said for blindfolding your statisticians)

[gustavolacerda.livejournal.com]

Authors can be (a) too reckless in claiming causality, or (b) too cowardly by beating around the bush because they are afraid of the word "cause". My impression is that in informal settings people tend to commit (a) more, and in scientific journals they tend to commit (b) more.

[omidia.livejournal.com]

amusingly, i read "causal" as "casual".

so i was confused as to why you were disagreeing with other stat'ans interpretations of their casual relationships... (laugh!!!)