MNAR Logistic Regression Mechanism

[sitaomin1994]

As described in the paper the first method for MNAR is using logistic regression with all other features as inputs, then mask data using MCAR. But in the code, I found it only has MNAR self-masking logistic regression, which does not take other features into account. We want MNAR not only self masking but also to have some correlation to other features, what do you suggest?

[BorisMuzellec]

Hi @sitaomin1994, there are two functions: MNAR_self_mask_logistic (which implements self-masking) and MNAR_mask_logistic (which implements the first mechanism you mentioned).

You can also check [this notebook]](https://rmisstastic.netlify.app/how-to/python/generate_html/how%20to%20generate%20missing%20values) from the r-miss-tastic website for more details.