2
$\begingroup$

I've created a logistic regression model in R using the glm function using a bank data and. I am able to get the AIC value, coefficients, percentage concordant, tied and discordant pairs in test and training data. But I have a validation data set where the status of the customer(good or bad) is not there as I have to check the performance of the model using this validation data. What should be my approach?

$\endgroup$

1 Answer 1

1
$\begingroup$

The concordant percentage cannot be computed without knowing the actuals (status of the customer, in your case). However, they can likely be estimated from the value computed on the training set. Under simplified assumptions the concordant percentage is apparently normally distributed around $d/\sqrt{2}$, where $d$ is Cohen's effect size. See Austin & Steyerberg.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.