01-25-2017 11:41 AM - edited 01-25-2017 11:41 AM
I am modeling a logistic regression model , with 14 explaination varibles , and 50,000 records.
Model fitting result:
All 14 variables are significant.
R-Square = 0.1927
Hosmer and Lemeshow Goodness-of-Fit Test
Chi-Square DF Pr > ChiSq
374.8494 8 <.0001
We can see the model only explain 19.27% of the data. Is this model useful at all ?
01-25-2017 02:26 PM
R squared isn't a good measure of model accuracy for LOGISTIC regression.
There is a pseudo-R squared value and other metrics that are more important. The documentation has examples or google search for evaluating LOGISTIC regression models.