Estimating Relationships Between Measures Derived from Responses to a Physical Activity Self-Efficacy Scale and Theoretically Relevant External Variables: A Monte Carlo Study

The objective of this study was to improve the validity of physical activity self-efficacy (PASE) measurement. To accomplish this objective, competing methodologies (i.e. traditional observed variable approach versus latent variable approach) for estimating external relationships for measures derived from responses to a PASE scale were systematically studied. Data were simulated using Monte Carlo experimental methods under conditions commonly observed in practice (e.g. N; one- versus two-dimensions; confirmatory versus exploratory models) to maximize potential implications for data analysis in practice. Parameter estimate outcomes (e.g. bias) were monitored. A one-dimensional latent variable model outperformed all other models (e.g. one-dimensional observed variable model) when the population model was one-dimensional. A two-dimensional exploratory latent variable model outperformed all other models (e.g. two-dimensional observed variable model) when the population model was two-dimensional. These findings provide a needed empirical basis for generally using a latent variable approach to assess a key validity guideline in PASE measurement.