Logistic Regression

#1
Hello
I am going to start building a Logistic Regression model based on medical data for the purpose of classification. My problem started with "Hosmer and Lemeshow test":shakehead and its insignificant result!, what should I do to solve this issue and what procedures should follow to make a correct analysis, rise the efficiency of the model and have the best fit for my model.
Thank you.
 

hlsmith

Omega Contributor
#2
Tell us more a out the model, sample size, proportion with outcome, number of covariate, format of covariates, HL. value etc.
 
#3
the sample size is 600 cases,the number of variables or covariates are 26, the overall percentage is 50%, the value of HL test (Chi Square statistic) is zero.
 
#4
Hi osamastat,

Hosmer-Lemeshow is a measure of calibration. Whilst it is important, there are other measures, such as discrimination, which which are important to describe when looking a a model. It may be that the model still has some useful application even if its calibration is not great, but it has excellent discrimination.

Can you tell us a bit more about what the model is for and how it has been developed? Are you trying to validate a model that has already been developed?