How to do model fit if the Hosmer Lemeshow GOF is less than 0.05

by Chen   Last Updated August 14, 2019 17:19 PM

I have a data set that the response variable is count data, with other predictive variables all categorical. I used the stepwise logistic regression to try to find the best model fit. I used the code as below in SAS:

proc phreg data=eyetest;
freq Dwell_Count;
class Task_Type Site AD_Type Ad_at_top /param=ref;
model Dwell_Count=Task_Type Site AD_Type Ad_at_top/ selection=stepwise slentry=0.25
slstay=0.15 details;
run;

But the result gave me the Hosmer Lemeshow GOF and Pearson Chi-square GOF both P-value less than 0.05. I am wondering what I can do for the next to find the best fit of the model.

Thank you very much!



Related Questions


Evaluating a logistic regression model

Updated October 24, 2017 19:19 PM


Vuong Test Equivalent for Logistic Regression

Updated April 13, 2017 10:19 AM