from your example it seems like multilevel analysis is the best you can do.
guo & zhao have written a nice articles, where you can find some
arguments for multilevel analysis:
Guo, Guang, and Haongxin Zhao. 2000. "Multilevel Modeling for Binary
Data." Annual Review of Sociology 26:441-462.
christian
Lee Woods schrieb:
> I have a data set of responses at the individual level that I wish to explain by measures of the areas in which these individuals live (measured at a census ward level.
>
> I have been reading about the statistical issues surround clustered (in this case geographically) data and the use of Multi-level models to overcome these problems. I am wondering whether it is neccessary in this case. The explanation I read as to why clustering is improtant to consider is that once clusted or grouped, repondents will tend to exhibit similar behaviours all else being equal. My question is that if I have theoretically described the areas in which the respondents live perfectly, do I need to have a multi-level model at all? Any suggestions would be very much welcome.
>
>
>
> Visiting Research Fellow
> Department of Civil Engineer
> University of Strathclyde
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
>
[hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
