Related variables at different levels

Welcome to the forum for MLwiN users. Feel free to post your question about MLwiN software here. The Centre for Multilevel Modelling take no responsibility for the accuracy of these posts, we are unable to monitor them closely. Do go ahead and post your question and thank you in advance if you find the time to post any answers!

Remember to check out our extensive software FAQs which may answer your question: http://www.bristol.ac.uk/cmm/software/s ... port-faqs/
Post Reply
MossyMcCollie
Posts: 10
Joined: Tue Oct 16, 2018 1:24 pm

Related variables at different levels

Post by MossyMcCollie » Wed Oct 31, 2018 10:26 am

Hi all
I have a 2-level model in which individuals (lower level 1) and nested within communities (higher level 2). The outcome measure is binary: whether or not an individual needs to access a particular social service.

One of the level-1 variables I need to include is ethnicity. It is considered very likely for various reasons that an individual's ethnicity will affect the probability that they will access the social service; for example, the impact of a family's ethnicity on the likelihood that they will be suffering economic hardship, be surrounded by (or be lacking) systems of informal support, be willing (or unwilling) to seek/accept help, and so on.

One of my colleagues has also requested that the same ethnicity information is additionally processed into a community (Level 2) variable (e.g. proportion of non-white individuals in the community). His logic is that there could be issues of context - so that the way, say, an Asian family is able to cope with personal/economic difficulties might be very different if they are in a virtually all-white community as opposed to a community with a high proportion of Asian families. Also, among other reasons, the community-level ethnicity profile could serve to proxy some important characteristics of the broader socio-cultural and economic environment of different communities.

I see his arguments, but I am somewhat hesitant about using the same data to create different variables at different levels of the model. I don't think I would generally use two variables derived from the same data in single-level regression model.

Does anyone have an opinion on whether my colleague's request would be valid?

Many thanks in anticipation

John

billb
Posts: 112
Joined: Fri May 21, 2010 1:21 pm

Re: Related variables at different levels

Post by billb » Tue Nov 06, 2018 10:24 am

Hi John,
It is quite standard to construct compositional variables in a multilevel model and add them to the model as contextual effects on top of the individual effects. If you think about our example datasets we talk about gender and school gender and one would use the gender variable to construct the school gender variable. In a regression there are cases where you use the same variable twice e.g. age and age^2 for example are derived from the same age variable.
Does that help?
Bill.

MossyMcCollie
Posts: 10
Joined: Tue Oct 16, 2018 1:24 pm

Re: Related variables at different levels

Post by MossyMcCollie » Sun Nov 11, 2018 1:21 pm

Yes, that is very helpful, Bill thank you very much. Apologies for the slow acknowledgement of your response: I have been offsite for a while.

Post Reply