0

I'm planning to work on some credit risk models using logistic regresson in R. Binary response. What all goodness of fit tests are to be known and WHAT PACKAGES are required for the same? Thank you. I found some help in this post but it is not exhaustive: What is the difference in what AIC and c-statistic (AUC) actually measure for model fit?

Community
  • 1
Srikanth Guhan
  • 106
  • 3
  • 13
  • Note that questions about specific software are off-topic here. Please see advice about software-specific questions in the Help Center. – Nick Cox Apr 28 '15 at 09:39

1 Answers1

2

You can use various pseudo-R statistics, which are based on deviance produced by the likelihood of full and restricted models. Deviance is produced by the glm function so you do not need any extra packages.

Hosmer-Lemeshow test can also be used to test global fit for logistic regression. It can produce result that link function used might not be right one.

http://artax.karlin.mff.cuni.cz/r-help/library/ResourceSelection/html/hoslem.test.html

Analyst
  • 2,527
  • 10
  • 11