You have quasi-complete or complete separation. Ideally you need more data - or to use an approach that adds more information (e.g., Bayesian regression - though other options exist).

Thank you so much for you answer! When I analyze my predictor variables independently (so I run two binomial logistic regression for each independent variable) the model is also significant but my Wald test is not. Is this a better way to report these findings?

No it's not. You don't have any information that suggests that your IVs will predict the DV. Given that you have nothing to report. See the attached paper for an example. Best wishes, David Booth

IMO, n=18 is far too small a sample to get anything very useful from an ordinary binary logistic regression model, even with just one explanatory variable. You could try "exact" logistic regression, but AFAIK, SPSS still has no procedure for that, as noted here:

• https://www.ibm.com/support/pages/exact-logistic-regression

So you would have to be willing & able to try using different software (e.g., Stata, SAS, R) if you wanted to try that. Here are some examples from the good folks at UCLA:

• https://stats.oarc.ucla.edu/stata/dae/exact-logistic-regression/
• https://stats.oarc.ucla.edu/sas/dae/exact-logistic-regression/
• https://stats.oarc.ucla.edu/r/dae/exact-logistic-regression/

Notice that Firth logistic regression is mentioned as another possibility. I'll leave you to do a search on that.

HTH.

Try to use R or Stata and run an ANOVA or multiple regression analysis. I hope you find this helpful.

I’m afraid to inform you that the interpretation of results of your analysis is incorrect. First, understand that logistic regression is a large sample statistical technique based upon maximum likelihood. It assumes a large sample; you don’t have that. Moreover the Wald’s test statistic is based upon a Chi-Square distribution, which is fundamentally a large sample statistic. The sample is too small to use that test statistic. The ratio of B to its standard error is not significant for each predictor. Nothing is here unfortunately!

I guess the sample size is too small for multifactor logistic regression.

A unifactor analysis would be more appropriate and fair for reporting.

You can search "rule of thumbs for logistic regression". This rule is still under discussion, but Your case is classic to apply it.

Good luck!

The output shown along with the original question shows that there are n = 18 observations, with 2 events and 16 non-events. That is not anywhere near enough events to fit a (useful) model, even with only one predictor variable, IMO.

You have only two fails. They are associated with all the explanatory variables. But the sample size is too small, so you do not have enough power to detect significant effects. Suggested solutions:

1- Expand the sample size.

2- Or, exact logistic regression.

3- Or, Bayesian approach.

Kind regards,

Wadie