notes.txt

allow rho to vary over time

hierarchical model for rho components

Find which three, say, series have the most overlap with this series.  Can figure rho for them? Then 2 stage?  Have to have country specific rho anyway.  Split cross-national series from country series?
> Still going to have to go through long-wide format to figure out which series have most overlap


need to follow basic model from 2008.  will need long-wide data, I think.


Problems with long-wide:
multiple 'identical' cy obs in SEDLAC Arg 2003:end net hh; CR 2010 net hh