Citation

O'Leary KD, Woodin EM. Clin. Psychol. Rev. 2005; 25(7): 877-894.

Affiliation

Psychology Department, Stony Brook University, Stony Brook, New York, 11794-2500, USA. K.D.OLeary@sunysb.edu

Copyright

(Copyright © 2005, Elsevier Publishing)

DOI

10.1016/j.cpr.2005.03.004

PMID

15921837

Abstract

Cross-sectional analyses from nationally-representative samples demonstrate significant age-related trends in partner aggression and problem drinking. Both behaviors are most prevalent in the early to mid-twenties and increasingly less common thereafter. Aggregate associations based on percentage of individuals displaying the behavior in each age range are dramatically stronger than those found when correlating individuals' ages and behavior. Multilevel modeling demonstrates that group-level effects do not mask associations found at the level of the individual for either problem drinking or partner aggression. An analysis of recent abstracts from psychology journals showed that issues of aggregate and individual data are rarely if ever discussed, and even well-known statistics books in psychology rarely discuss such issues. The interpretation of aggregate data will become increasing important as psychologists themselves, and in collaboration with epidemiologists and sociologists, have access to large data sets that allow for data aggregation. Both aggregate and individual analyses are valid, although they provide answers to different questions. Individual analyses are necessary for predicting individual behavior; aggregate analyses are useful in policy planning for large scale prevention and intervention. Strengths and limitations of cross-sectional community samples and aggregate data are also discussed.

Language: en