As an output of a logistic regression analysis I got an Estimate value of 0 (zero). Does this mean that the variable is useless in the model? Or what?
logit(P) = log(P / (1 - P)) = 13.458 - 0 Variable A - 0.106
Variable B - 0.004 variable C + 0.008 Variable D,
where P is Pr(y=1|x).
Just as an additional note: Generalized Linear Models always yield coefficient estimates. The fact that in your model the coefficient is zero does indeed mean that there is no reason to believe the variable has any impact on your dependent variable, but any other variable with a non-zero coefficient but a very large P-value has equally little 'effect'.
The coefficients with an estimate of exactly zero (and I by the way doubt that it is exactly zero, probably something like e-10 ;)) is not a separate class of coefficient estimates, so to speak!
Yes, a coefficient of zero for a variable means that that variable is not useful for predicting the log-odds of Y. The model’s prediction for the log-odds of Y does not depend on the value of the variable A.
