Common-item effects and the smallest space analysis of structure

Critically reviews procedures for the analysis of common-item effects on interscale structure. A study using smallest space analysis of such effects is described. It employed an original California Psychological Inventory scale intercorrelation matrix (with both common and unique items included) and a matrix based only on common-elements correlations between scales. Results indicate strong effects on structure of item overlap and suggest the generalization to smallest space analysis and related multidimensional scaling procedures of J. P. Guilford's 1952 admonition to avoid the factor analysis of test scales containing overlapping items. A distinction between common-item effects on magnitude of correlations and structure of intercorrelations between scales is discussed. It is argued that this distinction, aligned with a proposed research design, would aid in deciding when to use techniques such as factor analysis and smallest space analysis, including which technique is to be preferred under what conditions. (22 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved).