What If We Took Our Models Seriously? Estimating Latent Scores in Individuals

摘要

Researchers often argue that the structural models of the constructs they study are relevant to clinicians. Unfortunately, few clinicians are able to translate the mathematically precise relationships between latent constructs and observed scores into information that can be usefully applied to individuals. Typically this means that when a new structural model supplants a rival model, clinicians have only vague (and often incorrect) notions about how score interpretations should change. Fortunately, it is possible to estimate latent scores from observed scores in a rigorous manner. More important, paying attention to the confidence intervals around those estimates can assist clinicians' intuitions about what can (and cannot) be known with precision about a person's abilities. These methods are illustrated with structural models of the WISC-IV (Wechsler Intelligence Scale for Children-Fourth Edition). A free user-friendly spreadsheet that automates these procedures is available from the author.