I'm interested in general suggestions/discussions about how to answer questions that are requesting advice on how to perform an (arguably) ill-advised statistical practice. There are various degrees of consensus on what is "ill-advised", but two perennials that seem to be fairly well-agreed-upon are stepwise regression and tests of the null hypothesis of normality.

For example, in response to this "how do I do stepwise regression" question, I gave an answer that was basically "you probably shouldn't do this ... but here's how, if you really want to", whereas another user answered with an unconditional "don't do it". Going even farther in the "don't do it" direction, these questions could be closed as duplicates of the original "why you shouldn't do it" question ...


  • $\begingroup$ What do you think, Ben? Is this a duplicate or is there some distinguishing feature in your concern you would like to emphasize compared to the thread found by @StephanKolassa? $\endgroup$
    – whuber Mod
    Aug 24, 2018 at 21:28
  • $\begingroup$ yes, it's a duplicate. $\endgroup$
    – Ben Bolker
    Aug 24, 2018 at 21:47


Browse other questions tagged .