In:
Organizational Research Methods, SAGE Publications, Vol. 18, No. 4 ( 2015-10), p. 704-737
Abstract:
Management researchers often use consensus-based composition models to examine the antecedents and effects of higher-level constructs. Typically, researchers present three indices, r wg , ICC(1), and ICC(2), to demonstrate agreement and consistency among lower-level units when justifying aggregation. Nevertheless, researchers debate what values for these indices are sufficient. This study examines the distributional characteristics of ICCs and r wg values from three sources: the multilevel literature, a large multinational sample of student teams, and a large sample of randomly generated “pseudo teams.” Our results support existing cutoff criteria for ICCs but suggest that generally accepted values for r wg may, under certain circumstances, reflect pseudo-agreement (i.e., agreement observed among two raters not attributable to the same target). Thus, when there is minimal between-group variance (i.e., low ICCs), it is difficult to determine whether high r wg values reflect agreement or pseudo-agreement. Based on these findings, we provide recommendations to help researchers interpret aggregation indices.
Type of Medium:
Online Resource
ISSN:
1094-4281
,
1552-7425
DOI:
10.1177/1094428115582090
Language:
English
Publisher:
SAGE Publications
Publication Date:
2015
detail.hit.zdb_id:
2029600-9
SSG:
3,2
SSG:
3,4
Permalink