Forgive me if I may not have been standard in my formulation of the question.
I wish to analyze the effect of residential environmental factors on depression through linear regression. In the conceptualization of the questionnaire, there are many items in the dependent variable. Two of the questions are 1) whether there is a sports facility near the residence and 2) if so, what is the environment of that sports facility.
Now the problem is that question 2 is only asked if the answer to question 1 is "yes". So as it stands, question 2 will have a lot of missing values. However, in my linear regression, I want to study the effect of both question 1 and 2 on depression. In this case, how should I change the way the questionnaire is asked, or the way the model is constructed, so that both questions can be taken into account in the linear regression.
You will probably have to treat participants who answered the first question with "yes" vs. "no" as separate groups for the purposes of the regression analysis. Only for the "yes" group would it make sense to analyze the second question at all since that question does not apply to the "no" group.
Christian Geiser
Thank you for your answer, but I have one more question.
Because my model has not only these two independent variables, but also several independent variables, such as housing area and activity space area. So what should I do with these other independent variables if they diverge when modeling them separately? (For example, in the full-sample group, both housing area and activity space are significant. But in the group that answered yes to the first question, both housing square footage and activity space are no longer significant. Should I prioritize the results of the full sample model at this point?)
- Badges
- Science topic
- Similar topics
- Indexes