results with "e" in them? +inf?

[Sara Crowe]

Hello
I am conducting an exact logistic regression, but I get a coefficient of -2.14e-16. Does that mean something is wrong? I am guessing that it is not a number I can actually interpret and use?
And what does the +inf mean? is that also something gone wrong?

Would it be better if I asked stata to not estimate or condition out the constant? Would that bias my data?

Thank you so much in advance!

any_skin_s~s	Coefficient	Suff.	2*Pr(Suff.)	[95% conf.	interval]
sex	-2.14e-16	10	1.0000	-3.529724	4.479022
age	-1.334355*	9	0.2727	-inf	.977098
alcohol	-1.194763*	4	0.4000	-inf	1.59384
tobacco	1.458124*	7	0.2679	-1.049252	+inf
_cons	2.24643*	21	0.0165	.349224	+inf

[Maarten Buis]

2.14e-16 means \(2.14 \times 10^{-16} = 2.14 \times 0.00000000000000001 = 0.0000000000000000214\) (I may have miscounted the number of 0s), Lets just call that very small or basically 0.

+inf is + infinity and -inf is - infinity.

This does not look like a model I would trust, but leaving out the constant is much much worse. So don't do that.

I assume you are using exact logistic regression because you have perfect predictions, and a small sample size (the two are ofcourse related).