R help - Jul 2012 - cell blocking in log-linear models with a survey sample

Hello.  I hope someone can suggest a way to test log-linear models with 
cells
blocked when working with survey data. 

The tables  I am working with come from a multi-stage stratified sample. 
The 
command 'svyloglin' in the package 'survey' works well when I
want to
choose between models with two or three-way effects, but I haven't been 
able to figure out how to get proper p-values to compare models with and 
without blocked cells.

Thanks in advance for any suggestions.

Bob Arnold 
	[[alternative HTML version deleted]]