J. Haapasalo et al., Relative advantages of person- and variable-based approaches for predicting problem behaviors from kindergarten assessments, J QUANT CR, 16(2), 2000, pp. 145-168
Kindergarten teacher ratings of physical aggression, hyperactivity, inatten
tion, anxiety, and prosocial behavior were used to predict self-reported de
linquency, peer-rated social withdrawal, add school placement in preadolesc
ence (ages 10 to 12 years) in a large longitudinal sample of boys from low
socio-economic neighborhoods. Two analytic strategies were used: person-ori
ented and variable-oriented approaches. In the person approach, eight clust
ers, based on the kindergarten behaviors, were used to predict delinquency,
social withdrawal, arid school placement. In the variable approach, the ki
ndergarten behaviors were used as dimensions in logistic regressions. Famil
y adversity was used as the first predictor in both approaches; it signific
antly predicted all the outcomes in preadolescence. The results obtained us
ing the two approaches were partly in accordance. In both sets of results,
kindergarten teacher-rated externalizing behavior problems were most relate
d to later self-reported delinquency, and internalizing problems to peer-ra
ted social withdrawal. However, the person approach showed that all pattern
s of kindergarten behavior problems increased the risk for placement out of
an age-appropriate regular classroom, while only inattention and lack of p
rosocial behavior were the significant dimensional predictors of this negat
ive outcome according to the variable approach. Also, the person approach s
howed that the Multiproblem kindergarten boys had the highest percentage of
comorbidity of preadolescent problems. The ROC curves indicated that predi
ction of delinquency, social withdrawal, and school placement were equally
accurate using the cluster and variable approaches. Advantages and limits o
f both approaches are discussed with reference to their usefulness for clin
icians.