Models 3 and 4 were identical to models 1 and 2, respectively, except that they accounted for sibling fixed effects. For the full sample of all children in the mortality sample, low birth weight substantially increased the hazard ratio for mortality with and without sibling models (2.13 and 1.79, respectively), although it was not statistically significant. Among children from poor families, however, the hazard ratio was much larger and statistically significant (6.33 [P