Logistic Regression

[kenshin]

Can someone please tell me what logistic regression is? I know nothing of regression also (assuming that something like that exists). I read wikipedia, but couldn't understand much. what exactly the following line means:

"Logistic regression is a statistical regression model for binary dependent variables." Secondly,can someone please explaine each and every term of the equation given by wikipedia. Link is given below.

 

http://en.wikipedia.org/wiki/Logistic_regression

 

Thanks in advance

[Glider]

I can't explain the terms for you, but I can tell you a bit about regression.

 

Regression tests are based on correlation. They are a way of testing certain models for predictive power, and as such are more exploratory than some other statistical tests.

 

You take measures of a number of 'predictor' variables (i.e. variables you propose are predictive of another variable; the 'criterion' variable.

 

In linear regression, the criterion variable will be scale data (i.e. at least of ordinal level) where the values fall on a meaninful scale from low to high.

 

In logistic regression, the criterion variable is binary (i.e. dichotomous), for example, male, female, or drink, don't drink.

 

In either case, the regression tests each predictor variable against the criterion variable and will tell you whether the whole model itself is statistically significant, and whether (and to what degree) each individual variable is predictive of the criterion variable.

 

I hope that helps.

[Tartaglia]

A model of the form

 

ln(p/(1-p)) = a +bx +e

 

p = proportion/probability

a = intercept

b = gradient

e = error term

x is the independent variable

 

a and b are fitted to minimise errors

 

The model can be used with more than one variable too