This post has gotten 4 close votes.

I think the question is pretty clear, about the nature of AIC as model selection criteria for time series. A commenter requested that I include the code, and I did. But it's still getting close votes.

How can I improve it?


1 Answer 1


(Making my comment into an official answer.)

@StephanKolassa commented here explaining his concern with your question. The close reason is "In the meantime, I am voting to close as unclear". Three subsequent users have voted to close for the same reason, two of whom before you added the code, and one of whom after. I don't see that Stephan has revisited since his comments, but it's hard for me to tell. If it is closed & he thinks it has become answerable, I can reopen--just flag it.

Stephan revisted the thread and commented:

Thanks for adding the code; I have retracted my VTC. I have a suspicion and hope to find time later today to follow up on it. (Note that stepwise and approximation are only relevant for the first model, where I indeed get the same AIC as you do.)

I have voted to leave open in the close vote queue. My vote should be binding, meaning that it will no longer show up in the queue for others to vote to close on. You should be safe from closing now. Again, if it does close, just flag it and I can reopen it easily.

  • 2
    $\begingroup$ Thank you for that vote! This is an issue that I have been a bit unhappy about for a while: once a high-rep votes to close (VTC) and the thread ends up in the close-vote queue, I suspect a lot of other users more or less blindly also VTC. Even if the OP adds the requested clarification, and the original close-voter retracts their close vote. It's an unfortunate effect, but I don't know how to combat it. One could, of course, hold off on VTCing, but then I'd quickly have a long backlog of questions where I requested clarifications that I would need to revisit. Suboptimal. (ctd) $\endgroup$ Aug 8, 2018 at 13:44
  • 8
    $\begingroup$ (ctd) Also, my experience is that out of about ten questions where I request clarifications, only about one or two OPs actually do provide them. Which I'll admit pushes me towards requesting clarifications and simultaneously VTCing. Leading to the effect above. Then again, if a clarified question is closed, one can always nominate it for reopening, or flag it for a moderator's attention. $\endgroup$ Aug 8, 2018 at 13:50
  • 5
    $\begingroup$ @StephanKolassa, yeah, I've long thought that the system incentivizes voting to close. There isn't an ideal solution w/i the current setup. Your approach is the best you can do. $\endgroup$ Aug 8, 2018 at 14:31
  • $\begingroup$ @StephanKolassa , I'm going through the exact situation you described above here. I've already voted to undelete it, but I'm not sure if that's all I can do. Could you please provide some clarification on what you mean by "flag it for moderator's attention"? Also, feedback on the question is welcome. $\endgroup$
    – Eric Lino
    Aug 10, 2018 at 19:37
  • 2
    $\begingroup$ @EricLino: it's almost certainly possible to plot a fit line, by fixing all but one of your predictors to a specific value and then calling some predict function with a range of interesting values for the remaining predictor. You might be able to vary more than one predictor and plot an interaction. You might be able to extract standard errors to plot CIs. How exactly to do so is still off-topic here, which is why I assume the question is still closed (and auto-deleted). It is possible that the coefplot2 functionality was folded into coefplot, did you check? $\endgroup$ Aug 10, 2018 at 20:47

You must log in to answer this question.

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