6

I am writing up a logistic regression that looks at parenting factors (ie., parental depression, anxiety and stress) as predictors of the presence of a child anxiety disorder diagnosis.

What is the protocol re. reporting the bivariate correlations (accoridng to APA6- or even just generally)? Is it advised to include a table of the correlations before the logistic regression write-up? And if so, should this include the categorical DV- even though such correlations aren't very meaningful?

I have never written up logistic regression and am having trouble finding many examples.

Thanks for your help :)

Bernd Weiss
  • 7,044
  • 28
  • 40
JBurn
  • 61
  • 2
  • I encourage you to include them even if they only go in an online appendix. Our transparency should no longer be limited by the space constraints of a journal! In that spirit, I'd encourage you to consider going further and include a matrix of scatterplots, or grouped box-plots, etc. – Michael Bishop Oct 01 '12 at 19:00

2 Answers2

4

You might want to check the following papers which discuss how to report findings from logistic regression analysis:

From a meta-analytical point of view, it is always useful to report bivariate statistics. So, I always report them (put a table in the appendix), even if some of the variables are dichotomous. However, this also depends on the field you are working in.

Community
  • 1
Bernd Weiss
  • 7,044
  • 28
  • 40
2

I am pretty sure that APA6 makes no recommendation on this.

If this is for a journal, you should check with them. If they have online appendices, then @Bernd 's idea of putting the correlations in an appendix will almost surely work. If not .... well, in my reading in the social sciences and medicine, I rarely see the correlations reported. Page limits and all that.

If this is for a dissertation, it is almost certainly a good idea to put in the correlations.

Peter Flom
  • 94,055
  • 35
  • 143
  • 276