We began the analysis by comparing the health of young children in Mexican and U.S. households with and without migrant experience. Our analysis used children as the unit of observation and included all children in families in these communities. Because poor health may cluster in certain families and communities, we computed robust standard errors for the coefficients to account for multiple observations in households and communities. Using ordered probit methods that allow for the rank-ordering of outcomes without making assumptions about the intervals between categories, we then examined how household migration and network ties affect children's health with models that included dummies for the effects of immediate and extended ties, then successively added parent migration status and then control variables for individual and household characteristics. (Note that we focus only on the effects of immediate and extended ties—rather than include all the network measures—because these were not collinear with other variables in the models and because using other measures did not change the results described below.) Finally, because of sample-size limitations, we examined the extent to which network attributes affect the impact of family migration on child health using a series of interactions that specify these effects.