Dear Statalist,
I welcome suggestions for testing a linear regression model on a pooled series of cross-sectional surveys. This scenario might be familiar to some. Each survey cycle has a common multilevel structure: students are nested in schools, schools are nested in countries. About 40 countries participated 6 national surveys between 1994 and 2014. My research question relates to growth modelling because I want to pool these data and test synchronous and lagged effects of national unemployment on health. The purpose of using multiple survey cycles instead of one is to isolate period effects, similar to an accelerated longitudinal design.
Trouble is, countries are the repeated observations in the data, not schools or students. I don't want to aggregate health data to national averages so I can later examine some cross-level interactions. Can the xt tools in Stata be used for this sort of thing? for example,
xtmixed health c.year c.unempl || schoolID: || countryID:year, cov(unstr)
Frank Elgar
McGill University
I welcome suggestions for testing a linear regression model on a pooled series of cross-sectional surveys. This scenario might be familiar to some. Each survey cycle has a common multilevel structure: students are nested in schools, schools are nested in countries. About 40 countries participated 6 national surveys between 1994 and 2014. My research question relates to growth modelling because I want to pool these data and test synchronous and lagged effects of national unemployment on health. The purpose of using multiple survey cycles instead of one is to isolate period effects, similar to an accelerated longitudinal design.
Trouble is, countries are the repeated observations in the data, not schools or students. I don't want to aggregate health data to national averages so I can later examine some cross-level interactions. Can the xt tools in Stata be used for this sort of thing? for example,
xtmixed health c.year c.unempl || schoolID: || countryID:year, cov(unstr)
Frank Elgar
McGill University