Review of: Robust tests should be the default, not the backup. Round#1/Reviewer#1

hypothesis statistical significance tests, we should use robust tests as the default, not just when we detect assumption violations in classical tests.By robust tests, the author means tests that do not depend on assumptions about the distribution of residuals, or are more robust to extreme residual values.The author's case is that violations of residual Normality are in fact common, that detecting them is problematic, and that you gain reliable inference by using robust tests, without incurring a lot of cost.In short, it's better to be insured against violations when you don't need to be, than to be uninsured in the cases where they make a difference to the inference.The author addresses some of the recurrent concerns about robust tests (lack of power, non-availability for complicated designs) and provides simulated examples.All in all, the paper as it now is makes a case that is worth considering, and makes it well.The back and forth with reviewers sharpened the scope of the argument and the range of situations to which it applies.