Premium
The stability and coherence of aggregated and single‐item measures of antisocial behaviour
Author(s) -
Bukowski William M.,
FerberGoff Janet,
Newcomb Andrew F.
Publication year - 1990
Publication title -
british journal of social psychology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.855
H-Index - 98
eISSN - 2044-8309
pISSN - 0144-6665
DOI - 10.1111/j.2044-8309.1990.tb00897.x
Subject(s) - psychology , consistency (knowledge bases) , reliability (semiconductor) , statistics , coherence (philosophical gambling strategy) , observational error , social psychology , developmental psychology , econometrics , mathematics , power (physics) , physics , geometry , quantum mechanics
Aggregated and single‐item measures of antisocial behaviour taken from a one‐year longitudinal study of early adolescents ( N = 334) were examined according to their consistency across situations and the degree of coherence among them. To assess the extent to which differences between single‐item scores and aggregated scores could be attributed to differences in reliability, comparisons were made with scores that were either corrected or uncorrected for measurement error. Consistent with predictions derived from Rushton & Erdle (1987), both consistency and coherence increased as the number of items contributing to a score increased. However, as expected in light of the arguments of Campbell, Muncer & Bibel (1987), these increases appeared to be primarily due to higher reliability in the multiple item scores. Nevertheless, some increases in consistency and coherence remained after the effects of measurement error had been accounted for, thus indicating that the higher degree of consistency in aggregated scores cannot be due solely to their higher reliability. These results are important as they demonstrate that, relative to single‐item scores, aggregated scores provide a more consistent index of antisocial tendencies regardless of whether the effects of measurement error have been taken into account. Practical and theoretical issues regarding the use of single‐item and aggregated scores to represent behavioural constructs are discussed.