2

I applied a Hosmer-Lemeshow-Test (now named only goodness-of-fit?) provided by the RMS Package by Prof. Harrel. The p-values for my model were not significant, therefor applying the following link: Goodness-of-fit test in Logistic regression; which 'fit' do we want to test?

I fail to reject the Null-Hypothesis, therefor I cannot reject the model if that is correct.

However while applying also a Pseudo-R² measure with extremly low levels (from a different package, but the one from the RMS-package suggest the same) aswell as plotting the ROC and calculating the AUC, I ask myself: How to interpret the Hosmer-Lemeshow-Test?

Is it right to state: While it is not possible to reject the model, there is plenty of room for improving it? Can I state something like: Even though the model isn't significantly or logically false it doesn't represent the data appropiately? What statement can I really conduct from this outcome?

I am still a nooby in this platform, please don't pick on me again.

Nuke
  • 87
  • 11
  • In this post https://stats.stackexchange.com/questions/35422/validation-of-logistic-regression-goodness-of-fit-pearson/35427#35427 it is stated that Hosmer-Lemeshow test is now considered obsolete – kjetil b halvorsen Aug 01 '17 at 10:25
  • I know, but it is the replacement-test for Hosmer-Lemeshow which is now named goodness-of-fit test? The author of this test in rms package is the same author of this post, naming hosmer-lemeshow obsulete. does it mean the test is obsolete or does it mean also the new adoptions of it are obsolete. – Nuke Aug 02 '17 at 12:01
  • 2
    I am sorry about the (old) downvote. We cannot control voting; it's up to community members to decide which questions might need more research or clarification. *I* think this is a well-formulated question, so I have upvoted it. But I only get one vote... . – whuber Oct 09 '18 at 16:01
  • Package `rsm` does not compute the Hosmer-Lemeshow tests, but can compute some better tests. See functions (in `rsm`) `val.prob` and `residuals.lrm`. – kjetil b halvorsen May 14 '20 at 12:06

0 Answers0