Predicting Placement in Foster Care
Thomas P. McDonald, John Poertner, Gardenia Harris · Journal of Social Service Research · 2001
In this paper we explore the use of neural network analysis (NNA) as an alternative to logistic regression to predict which children with a founded (indicated) child abuse/neglect report will be subsequently placed in foster care. The main advantages of NNA are that it is a nonparametric technique requiring no assumptions of normality that can readily accommodate both linear and nonlinear relationships and interactions without prior specification by the researcher. The two techniques were found to yield similar classification results for these data; however, NNA provides unique capabilities in analyzing and displaying interactions in predictor variables that may make it more useful for data mining.