2
$\begingroup$

My question1 has been marked "duplicate"2-4 and closed. I believe it's a misunderstanding, since the other question my question has been associated with is different. The other question asks why a given variable becomes non-significant when included in a more complex model. My question asks why a given variable becomes significant when included in a more complex model. That's exactly the opposite, therefore the answers given in the other question (which I read before posting) don't help me.

Is there something I can do to re-open the question?

  1. Linear regression: why is a variable not significant on its own, but significant together with other variables
  2. Significant predictors become non-significant in multiple logistic regression
  3. How can adding a 2nd IV make the 1st IV significant?
  4. Is there a difference between 'controlling for' and 'ignoring' other variables in multiple regression?
| |
$\endgroup$
  • 6
    $\begingroup$ As you might imagine, such questions are fundamental and surely, after this site has operated almost ten years, will have been formulated and answered in many ways. Your first thought, then, should be to search our site for those answers. One that has worked well for this particular issue is regression significant change. $\endgroup$ – whuber♦ Jun 4 at 12:59
  • $\begingroup$ @whuber Thanks, I understand that, and did search the site before posting. However I couldn't really find an explanation that fits my situation (and one that I can understand as non-statistician). Therefore posted as separate question, with specific sample data, in hope to get more specific ideas or suggestions to understand the problem. If that's too much to ask per the policy of CrossValidated that's fair enough, then I withdraw my request. (P.S. link #3 is actually the most similar one to my question, perhaps this should be associated) $\endgroup$ – Michael Dorman Jun 4 at 13:27
  • $\begingroup$ As you can see, we had already a number of other questions that answered very similar questions. Three of them are now marked as duplicates of yours. Is there any reason why they don't answer your question? $\endgroup$ – Tim♦ Jun 4 at 13:34
  • $\begingroup$ @whuber I now see that my question does appear on the website, it's just marked as "duplicate". Initially it seemed that trying to enter my question automatically redirects to the associated older question, which is why I posted here to try to revert the situation. If the question still appears then it's fine, thanks for correcting this! $\endgroup$ – Michael Dorman Jun 4 at 13:36
  • $\begingroup$ @whuber answer #3 does give possible reasons how this may happen, but my question (two bullet points at the end) were actually more practical-oriented: Can I still conclude X1 affects Y? What kind of tests and diagnostics should I run and report in such situation? $\endgroup$ – Michael Dorman Jun 4 at 13:40
  • 2
    $\begingroup$ @MichaelDorman when you add or remove variables from the model, you assume that this would impact the model, otherwise why would you do this? Controlling on other variables, some variables may have different impact on the model than before. That's what the other threads discuss. $\endgroup$ – Tim♦ Jun 4 at 13:44
  • 3
    $\begingroup$ I appreciate your constructive engagement with the issues (+1). Some actions that often help are (a) if possible, shorten the text; (b) explicitly discuss the relevance (or lack thereof) of apparent duplicates on the site (or even elsewhere); (c) focus the post on a single clear question. Because the change of sign or significance of variables in regression has been so extensively discussed, the challenge for you is to demonstrate how your question is different. (I suspect it might be, which is why I'm writing these suggestions.) $\endgroup$ – whuber♦ Jun 4 at 13:46
  • 1
    $\begingroup$ @whuber Thanks! I edited the question to clarify that I'm looking for practical advice rather than theoretical considerations on why this happens. $\endgroup$ – Michael Dorman Jun 4 at 13:53
7
$\begingroup$

I think you're right that the question at the bottom of your post ("What should be the conclusion regarding the effect of b on value?") is OK. In light of that, I'm willing to reopen (and answer) it.

For future reference, that question doesn't match the question in the title of your post very well ("why is a variable not significant on its own, but significant together with other variables"). Moreover, the title question is much more salient, and is a frequently asked question—that's why it was closed as a duplicate (rightly, in my view: I read it before adding additional duplicate targets and interpreted the question at the bottom in light of the title). It is your comments here, and your edit to your post, that clarify what the real question is. For the record, those clarifications are the way to productively deal with having your post closed as a duplicate if you disagree.

| |
$\endgroup$
  • $\begingroup$ I appreciate it, thank you! $\endgroup$ – Michael Dorman Jun 4 at 18:39

You must log in to answer this question.

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