Logistic regression analyses were used to establish the most robust set of risk factors that would best predict borderline/clinical levels of problem behavior (i.e., a t score at or above 60 on the Child Behavior Checklist Total Problem scale) of kindergarten and first-grade children at risk for emotional and behavioral disorders. Results showed that among the 11 risk factor domains considered, 5 were most predictive of borderline/clinical levels of problem behavior: externalizing behavior pattern, internalizing behavior pattern, early childhood child maladjustment, family functioning, and maternal depression. Within these 5 domains, the most robust set of individual risk factors were difficult child (i.e., temperament, parent management skills, interaction between temperament and parent management skills), destroys own toys, and maternal depression. Results, limitations, future research, and implications are discussed. (Contains 3 tables.)