Interpreting Logistic Regression

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Interpreting Logistic Regression
Message posted by Xavier Olave on November 8, 1999 at 12:00 AM (ET)

I wish to learn how to interpret the coefficients in the multivariate logistic regression model (i.e., what does the positive or negative coefficient signify in the Logistic Regression Equation?)

Thanks.

Xavier Olave
Chicago, IL

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Interpreting Logistic Regression
Message posted by JG on November 8, 1999 at 12:00 AM (ET)

I presume you have a model of y = a*(x)**b and that you fitted log(y) = log(a)+ b*log(x) to make it a linear regression. It should now be obvious what log(a) and b mean whether they are positive or negative.
The big problem is that if you assume that the error terms of log(y) are i.i. and normally distributed is this reasonable and what does it say about y.

Re: Interpreting Logistic Regression
Message posted by Jack Tomsky on November 9, 1999 at 12:00 AM (ET)

In a multivariate logistic regression model, the dependent variable Y is either zero or one, representing the absence or presence of an event.

The model is Prob(Y=1) = z/(1+z), where z = exp[A + Sum(Bi*xi)]. A Bi of zero indicates that the factor xi has no effect on the probability of Y.

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.