1
$\begingroup$

I am trying to figure out the best way to report the results of a logistic regression in an APA paper. My understanding is that the odds ratio is the most important for interpretation so I don't think I should report the Beta. For significance, do I state the Wald like I would in a comparison (e.g. t(N-2) = #.##, p < .##). Thanks.

$\endgroup$
  • $\begingroup$ This should probably go to a different SE site (maybe Academia or Psychology & Neuroscience, I don't know for sure). $\endgroup$ – gung - Reinstate Monica Jul 16 '16 at 15:28
  • 1
    $\begingroup$ At any rate, a Wald statistic is not distributed as t; it is a standard normal. I would think an acceptable form would be "exp(beta) = #.##, CI = (#.##, #.##), p = #.###". $\endgroup$ – gung - Reinstate Monica Jul 16 '16 at 15:30
  • 1
    $\begingroup$ We do have a "reporting" tag, and I think generally reporting results is considered to be on-topic here. $\endgroup$ – Silverfish Jul 16 '16 at 17:01
1
$\begingroup$

If you have hundreds of records in the run, then the standard normal distribution $N(0,1)$ approximates the $t_{\nu}$, and you could use Z-scores for the test statistic, that is $Z_j=\beta_j/s.e.(\beta_j)$. If $\textrm{abs}(Z_j) \geq 1.96$, then the $OR$ is significantly different from unity. By the way, the $OR$ for variable $j$ is simply $\exp\{\beta_j\}$. We almost always use $Z$ when reporting everything, for example: "results were significant $(Z=2.01, P=0.04)$." If you have small sample sizes then you are constrained to using $t$-variates for $\nu=1$ d.f.. The Wald $\chi^2_{\nu}$ statistic is merely $Z^2$, since for a 1 d.f. test, $1.96^2=3.84$.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.