Statistics are never true (Score 1)

All this talk about truth...

Statistical findings are never true, even when they are essentially accurate. They are guesses about properties of populations that are themselves hypothetical. The problem is that the scientific community likes to portray such fundamental methodological weakness as if it were precise, replicable, theoretically informed truth. Statistical findings are, at best, weak empirical indicators that only strengthen when underlying mechanisms are understood, at whch point the statistics themselves become superfluous.

Peer review cannot overcome the problems with theoretically weak, poorly executed and interpreted statistical research--even when population parameter estimates are accurate. The peers are themselves too committed to the paper production system. If the empirical goal of identifying important problems and then tracking down underlying mecanisms is secondary to paper production--which too often is more about making journal editors feel comfortable than about the serious debate needed for scientiic advance--then statistical science becomes mostly a fashion show. I think the author of the article is onto this and, yes, many weak scientists probably should feel threatened and probably would not be happy with this analysis. They have no sensible argument against it. Paper production is only the goal when you are not advancing as a science. I would venture to say that the more papers in a field by more people blindly espousing the statistical empirical ideology --while never actually following a problem to a more definitive conclusion-- the more likely we have a science that is not advancing, and probably cannot advance. Reporting scientific advance should be the goal, not numbers of papers produced. But papers produced is the coin of the realm in what is presented as legitimate science. Thus, much of this corpus of "scientific literature" is trash.

I am very glad the author of the paper being discussed has raised these important questions.