Question 288:

Asked on July 24, 2008

Tags: Newton-Raphson , Logistic Regression , Maximum Likelihood

1

Answer:

No answer provided yet.

For this data-set I'd use a binary logistic regression analysis. This type of regression test is appropriate when your dependent variable is binary (yes/no) and is coded as a dummy variable with a 1 or 0. The logistic regression can handle qualitative predictors, also coded as dummy variables as well as continuous predictors. The logistic regression will compute the log-odds of the dependent variable while holding each predictor constant, just like as in ordinary regression analysis. The computations are done using Maximum Likelihood , based on a Newton Raphson algorithm and is not closed form as in least-squares regression.

See Alan Agresti's Introduction to Categorical Data Analysis, 1998 for a good resource.