For ordinary statistical models (on correlation), we can always think of some criterion of prediction accuracy to compare models(mean square error, AUC curve, concordance index, etc).

However for causal models...I read the unpublished manuscript of Hernan& Robins...it seems that the message they are conveying from Part II is "since these models give pretty close estimate, they must be all good"... https://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/.

Are there any way to give preferences to causal models, for instance, that one is superior to another?


put on hold as unclear what you're asking by Carl, Michael Chernick, jbowman, mkt, mdewey 2 days ago

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • $\begingroup$ There are two types of causality. One is probable causality, the other is deterministic causation. The latter has a mechanistic chain of cause and effect. In probabilistic causation, a mechanism for causation may be unknown. For example, there is a qualitative difference between "a drought can cause a famine," and "smoking can cause cancer." In the former case, we can easily picture that rainfall can cause crops to grow, in the latter, non-smoking implies nothing particular, and only the former is clearly deterministic. $\endgroup$ – Carl Jan 15 at 5:47
  • $\begingroup$ Sure. I mean probable casuality of course or I might ask in a different forum lol. Though, my question is this: We know that smoking cause cancer not the reverse. But in some settings, the direction of causality is not known by common sense and we have observational data. The data structure is not time series so we can not utilize the fact that cause comes first. In this case, if one causal procedure says A causes B and the other says B causes A or they have a common cause, what can we do? $\endgroup$ – failedstatistician Jan 16 at 17:12
  • $\begingroup$ Superior how: more causal, more physically correct, more precise/accurate within the range of data, more predictive of withheld data and extrapolates or back-extrapolates better, is more predictive of a parameter that is implied by the model, but is not a simple modelling parameter or what? Here is some preliminary help for causality stats.stackexchange.com/q/387244/99274, but that does not even begin to answer your question, because you need to finish asking it. $\endgroup$ – Carl 2 days ago
  • $\begingroup$ Please quote the exact text from which you have abstracted the parenthetical quote. It is unclear what this impression is gleamed from. $\endgroup$ – Carl 2 days ago

Browse other questions tagged or ask your own question.