Cutoff value of a continuous (discrete) variable - prediction

[Stefanos Demertzis]

Dear Statalists,

I hit a mental roadblock. I have a binary dependent variable (a medical diagnosis) and a small series of independent variables. I fitted a model of logistic regression, in which 3 discrete variables (d1, d2, d3) influence the the dependent variable significantly. From visual inspection of histograms (and some matter knowledge) I see a specific value of d1 (a specific ratio) being like a cutoff in turning the dependent variable from 0 to 1. Is there a statistical method to prove it?

[Maarten Buis]

Maybe something like this: http://maartenbuis.nl/publications/leaps.html

[Stefanos Demertzis]

Thank you Maarten! Very useful, especially the graphs! I tried something very similar but abandoned it (too early!). With your precious input I'll get back to this more motivated!! Thank you!