Likelihood ratio test in Logistic Regression

umvue · April 17th, 2009, 2:24 am

Suppose I have a case control study with the following numbers:x 0 1 2case 89 369 342control 56 250 266The logistic regression looks like this and x can only take the value of 0, 1 and 2logit(p) = alpha + beta*xThe answer for the likelihood ratio test for beta=0 is approximately 1.991But how do I calculate it? Thanks a lot!

umvue · April 24th, 2009, 8:41 pm

I get it now. You have to find the difference between deviance and null deviance.There doesn't appear to be a simple formula based on counts that can do the trick. If someone knows, please let me know.BTW, is logistic regression used in finance?

va1210 · April 26th, 2009, 6:11 am

QuoteOriginally posted by: umvueBTW, is logistic regression used in finance?Sure, it is quite commonly used in credit scoring when predicting probabilities of default.

quantmeh · April 26th, 2009, 12:25 pm

QuoteOriginally posted by: umvueIBTW, is logistic regression used in finance?we use it. think of prob of default. the default event is 1 or 0. PD is not measurable, obviously, but u can model it with logit