Psychometric and item response theory methods have long been used to estimate scores or latent traits. Although traditionally used for testing, these methods are applicable to other problems to reduce dimensionality or provide aggregate measures. This presentation briefly introduces models from item response theory and psychometrics, motivated by two real research applications. First creating a composite from biological markers, and a second assessing individuals’ levels of perceived control and risk from cancer. Measurement models are presented first as latent variable (factor) models and second as mixed models. Demonstrations show classical methods using maximum likelihood estimation as well as Bayesian methods for both mixed models and latent variable models.
Joshua Wiley has been working as a statistical consultant for the past two years, and is currently a senior analyst at Elkhart Group Ltd. He is also completing a doctorate degree in health psychology where he studies how individual differences and mechanisms linking psychological and physiological systems. Particularly how modern analytics can integrate different levels of analysis from the micro biological level to more macro psychological and social ones.