The following question (let's call it A) Unequal sample size one way ANOVA has an accepted, upvoted answer. It is also marked as a duplicate of the newer question (B) Unequal Sample Sizes One-Way ANOVA. B has 2 lower-quality answers, both with net negative scores.

This seems amiss. I'd like to flag a third question as a duplicate, and it makes more sense to point to A, given the quality of the answer. Rather than having a chain of duplicates, is there a better solution here?

Swapping the direction of duplication (if this were possible) may not be suitable since B has multiple parts to it, not all of which are addressed in A (or any existing answer in B, for that matter).

| |
  • 2
    $\begingroup$ As you said, thread B is low quality overall. It's also pretty old and the OP is inactive. It does not really matter if the question in B is not fully addressed in A. It still makes sense to make B a duplicate of A. So I suggest you flag B for mod attention and ask to revert the duplication direction. $\endgroup$ – amoeba Aug 21 '18 at 19:09
  • 3
    $\begingroup$ See also stats.stackexchange.com/questions/122974. I think it's a better target. $\endgroup$ – amoeba Aug 21 '18 at 19:50
  • $\begingroup$ @amoeba Thanks! Good solutions. $\endgroup$ – mkt - Reinstate Monica Aug 22 '18 at 5:38
  • $\begingroup$ And which Q did you want to close as a duplicate ("I'd like to flag a third question as a duplicate")? $\endgroup$ – amoeba Aug 22 '18 at 9:48
  • $\begingroup$ @amoeba Whoops, I'll need to find that one again... Will flag when I locate it. $\endgroup$ – mkt - Reinstate Monica Aug 22 '18 at 11:21

It is possible for moderators to switch which thread is considered the master and which is the duplicate. It isn't necessarily the case that the oldest needs to be the master—the thread with the best answers can be (see: If question A came before question B how can A be a duplicate of B?, and Many (historical) duplicate questions - how to pick a “canonical” question?). However, a legitimate issue is whether a moderator should override the status quo based on a single flag. A reasonable approach is that if there is any ambiguity, the question of which would serve as a better master could be raised on meta.CV.

In this case, the two threads clearly overlap, but also both have nonoverlapping elements. My guess is that the most conservative option is for neither to be marked as an exact duplicate. I think little is lost by this.

Update: I switched "A" to be a duplicate of @amoeba's suggested thread. It's clearly a better target. In addition, I closed "B" as a duplicate of a list of threads that address the various questions raised, starting again with @amoeba's target, but also threads pertaining to the assessment of heteroscedasticity and alternative testing strategies in the presence of heteroscedasticity.

| |
  • 6
    $\begingroup$ Another option - particularly if there's at least one good answer among the poor ones, or if the one with poor answers is actually the more canonical question - is to merge. It only requires that the answers for the one being merged will serve as good answers to the target question. $\endgroup$ – Glen_b Aug 21 '18 at 23:07
  • 1
    $\begingroup$ Thanks, gung & @Glen_b. Useful to know that it's possible to mark a question as a duplicate of multiple threads. $\endgroup$ – mkt - Reinstate Monica Aug 22 '18 at 5:39

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .