(2015-08-12) Hannan, Michael T; Rubinson, Richard; Warren, Jean Tuttle
The authors describe and analyze some issues in understanding causality from panel designs. They focus on complications that arise when multivariate panel models are measured with either random or systematic errors. The analysis is illustrated with panel from the U.N. of education and economic data from 96 countries. They conclude that new statistics, to be developed or imported from other disciplines are needed to deal with measurement error in substantive panel data.