1
$\begingroup$

Greetings statistics experts, I am having a try with the kaggle titanic dataset and am wondering what to do with the residuals after fitting models. In the case of linear regression you can look at a residual plot and see if there is any discernible pattern, but this is not so straightforward with logistic regression.

How can I improve on my model.

Thank you.

$\endgroup$
2
$\begingroup$

There are partial residuals you can calculate for a logistic model, and these can be used to check the transformations of the variables. But this is a very subjective and indirect method compared to just modeling flexible predictor effects in the first place, i.e., by using restricted cubic splines. A detailed case study using the Titanic is given in the Handouts entry in http://biostat.mc.vanderbilt.edu/rms.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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