IBM Support

Is there a general rule for what is a good value of the -2 log likelihood for a logistic regression model?

Troubleshooting


Problem

I'm running a logistic regression model and am trying to understand what the -2 log likelihood value means. Is there a general guideline or rule for what that value should be for a good model?

Resolving The Problem

There is no guideline or rule for what the -2 log likelihood value should be for a good fitting model, as that number is sample size dependent. If the number being reported is -2 times the kernel of the log likelihood, as is the case in SPSS LOGISTIC REGRESSION, then a perfect fitting model would have a value of 0. (If the value printed is -2 times the full log likelihood value, as is the default in the NOMREG and PLUM procedures, the value would be a sample dependent constant rather than 0; see Technote 1476887). How far above 0 the -2 LL value will be depends upon the sample size as well as the degree of residual variation in the observations when compared with the predicted values, so there is no absolute way to judge these numbers.

[{"Product":{"code":"SSLVMB","label":"IBM SPSS Statistics"},"Business Unit":{"code":"BU059","label":"IBM Software w\/o TPS"},"Component":"Not Applicable","Platform":[{"code":"PF025","label":"Platform Independent"}],"Version":"Not Applicable","Edition":"","Line of Business":{"code":"LOB10","label":"Data and AI"}}]

Historical Number

49420

Document Information

Modified date:
16 April 2020

UID

swg21480582

Page Feedback