A second distinction separates models into those with single indicators and those with multiple indicators for each trait-method unit. In single-indicator approaches like the traditional multitrait-multimethod (MTMM) matrix of Campbell and Fiske (1959), there is only one indicator for a trait-method unit (e.g., one self-report item measuring extraversión). Multiple-indicator approaches require at least two indicators for each trait-method unit (e.g., several self-report items measuring extraversión). Multiple-indicator approaches have the advantage that unsystematic measurement error can be more appropriately separated from systematic method-specific effects, and that the generalizability of method effects can be more adequately analyzed (Eid et al., 2003; Marsh & Hocevar, 1988).
Was this article helpful?