3
$\begingroup$

For a particular question in a survey, respondents have been asked to select all that apply (i.e. say from a list of books they have read).

I'm wondering if anyone knows how I would be able to build a model that predicts which book they have read based on demographic variables obtained in the survey?

I know it would be possible to construct a model like this using Multinomial Logistic Regression if they were asked to select a single book (i.e. perhaps selecting the last book from the list they have read, if any).

Any suggestions/help would be greatly appreciated.

Thanks!

$\endgroup$

1 Answer 1

1
$\begingroup$

I think that in this case, you really have multiple dependent variables captured by the "select all that apply" question.

The way that I would probably approach it would be to create a model for each of the possible responses to the select all that apply question, effectively ending up with a prediction model for each of the possible responses.

Alternatively, you could also create models that capture combinations of the "select all that apply" questions, something of the form "the respondent selected at least one of the first three options is predicted by these explanatory variables."

It really comes down to what your research questions are.

$\endgroup$
1
  • $\begingroup$ Thanks for your comment Mike! Yea, I wasn't sure how I would be able to somehow combine all the cases into a single model. Your idea sounds good. Thanks again! $\endgroup$
    – Jordan
    Commented Dec 12, 2014 at 19:08

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.