1
$\begingroup$

I have been looking for an answer to my question here stackExchange and by googling, but although some have had similar questions, I am not sure, I have found the right answer.

I am running a logistic regression in which I am analyzing the probability of a child being placed in foster care. Some of my independent variables are related to the childs parents - for example income and education. However, for some of the children a mother or a father is not present, and I therefore don't know their income, education etc. My question is, should I code the income for a missing parent to be zero or should I just leave it blank? I should note that for both the mother and the father, I have already created a binary variable that tells if the parent is unknown.

If I leave the income for a missing father blank, does that mean, that the child with the missing father and missing income, isn't a part of my regression?

$\endgroup$
1
$\begingroup$

My question is, should I code the income for a missing parent to be zero or should I just leave it blank?

If the data is missing because there is no information on that parent's income, then imputing the missing data with 0s creates a bias. The reality is that the parents likely made some money.

You could either: Drop variables with missing information or use an imputation method (maybe MICE or something else). If you choose this route, I would research what the typical method of comparing regressions with and without imputation is.

| cite | improve this answer | |
$\endgroup$

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.