Being Inconsistent About Consistency: When Coefficient Alpha Does and Doesn't Matter

Abstract

One of the central tenets of classical test theory is that scales should have a high degree of internal consistency, as evidenced by Cronbach's a, the mean interitem correlation, and a strong first component. However, there are many instances in which this rule does not apply. Following Bollen and Lennox (1991), I differentiate between questionnaires such as anxiety or depression inventories, which are composed of items that are manifestations of an underlying hypothetical construct (i.e., where the items are called effect indicators) and those such as Scale 6 of the Minnesota Multiphasic Personality Inventory (Hathaway & McKinley, 1943) and ones used to tap quality of life or activities of daily living in which the items or subscales themselves define the construct (these items are called causal indicators). Questionnaires of the first sort, which are referred to as scales in this article, meet the criteria of classical test theory, whereas the second type, which are called indexes here, do not. I discuss the implications of this difference for how items are selected, the relationship among the items, and the statistics that should and should not be used in establishing the reliability of the scale or index.

A confiabilidade de um questionário avaliando as práticas diárias de trabalhadores envolvidos com a saúde mental da comunidade

“…Cronbach's alpha was 0.76. This value meets the standards established by Streiner (2003), (12) who suggests that the coefficient's values must be >0.7 to be reliable.…”

Reliability and construct validity of the Online Cognition Scale in the Portuguese (Brazil) version (OCS-BR)

“…The reliability evaluation involved an estimation of the internal consistence, Cronbach's Alpha, that must vary to being adequate between 0.8-0.9 23 . The instrument timestability was verified through the intraclass correlation coefficient (ICC) which must present scores ≥ 0.7 24 .…”

Bandwidth and Fidelity on the NEO-Five Factor Inventory: Replicability and Reliability of Item Cluster Subcomponents

“…Streiner (2003b), among others, emphasized the well-known fact that is strongly affected the number of items in a scale and, like Cortina (1993), cautioned against blind use of to evaluate an instrument's reliability without consideration for factors that affect (Streiner, 2003a). For instance, even though is intended for unidimensional scales, these results clearly show that was high for the multifactorial NEO-FFI domains.…”

“…Whereas and were intended for use with congeneric item sets, , which entails no such assumption, may be regarded as a more appropriate reliability estimate for the domains. However, domain s af-firmed the well-known psychometric fact that may be high in multidimensional scales (Streiner, 2003a(Streiner, , 2003b and thus should be not be regarded as a measure of "homogeneity" (Cortina, 1993 Table 4 presents the subscale intercorrelations, with the young adults below the diagonal, the middle-aged sample above, and correlations among subcomponents within each domain bolded in corresponding off-diagonal blocks. The pattern for both samples suggested moderate to strong correlation among subcomponents within a domain and much smaller correlations between item clusters from different domains.…”