Stepwise logistic regression (backwards LR)

I have done a backwards LR stepwise logistic regression with a small data set of 22 cases and have a few questions. The full model had four variables and the final model had two significant variables.

1). Is it necessary to put the regression equation in a scientific paper? This was an exploratory study.
2). In the final model, can variable 1 (p = .01, Odds ratio 1.534) be said to be more important than variable 2 (p = .03, Odds ratio 1.006)?
3). In a summary table of parameter estimates, is it necessary to put the constants in the table?

Thanks in advance for your kind help.

Edit: I uploaded the data set and SPSS analysis
Last edited:


New Member
Hi dmausel

Allow me to give you following inexperienced points of view.

1) I would say it is necessary to put the regression equation in the paper. This does not have to be explicitly, though. You can make a table with the estimates, from which the equation is obvious.
2) It depends on the variables. If they are binary, then I'd say yes - the first one has a larger effect than the second one. If one of them, or both, are ordinal, then it depends on how they are measured. If the first one is age in decades and the second one is height in nanometers, for example, the height one is probably the most "important".
3) Yes, I would say it is necessary, as you have to know the intercept in order to calculate the expected probabilities.

Can I ask you (in relation to my own thread on the forum) if you have performed the deviance test and calculated Nagelkerkes R^2? If so, what was the results?