Přehled o publikaci

Psychological questionnaires, often designed to measure latent constructs, frequently exhibit spurious multidimensionality, particularly when incorporating a mix of regular and reversed items. Existing explanations attribute this multidimensionality to construct-irrelevant factors (such as) or consider it a spurious outcome resulting from imperfections of the measurement model. This presentation focuses on the phenomenon of spurious multidimensionality, a unifying account of why it can arise. We posit that a primary cause of spurious dimensionality lies in an incorrectly specified item response function, where the empirical and model-implied relationships between a latent variable and its indicators differ. Then, when multiple items share a similar pattern of these discrepancies (residuals), a unidimensional model will not fit the data, necessitating the introduction of additional factors to account for the residual relationships between items. An empirical study supported our predictions. Items exhibiting a shared pattern of residuals tended to form an additional factor. Importantly, this factor still contained construct-relevant variance. Consequently, our findings underscore the importance of scrutinizing whether the relationship between a latent variable and its indicators is modeled correctly to avoid spurious multidimensionality. Otherwise, there is always a risk that spurious factors will be interpreted substantively.