Foster care reunification: An exploration of non-linear hierarchical modeling

Putnam-Hornstein, E. & Shaw, T. (2011). Foster care reunification: An exploration of non-linear hierarchical modeling. Children and Youth Services Review, 33, 705-714. https://doi.org/10.1016/j.childyouth.2010.11.010

This methodological article explains how to interpret non-linear multilevel models in child welfare research. In linear models, estimates mean much the same as in standard models, but in non-linear models adding a random intercept makes estimates “subject-specific” rather than “population-averaged”, and the two can differ substantially. Using California administrative data, the authors illustrate the difference with two examples that estimate the odds of family reunification using both standard logistic regression and random intercept models.