Correcting the bias of the Root Mean Squared Error of Approximation under missing data

Missing data are ubiquitous in psychological research. They may come about as an unwanted result of coding or computer error, participants' non-response absence, missing values be intentional, planned designs. We discuss the effects on χ²-based goodness-of-fit indices Structural Equation Modeling (SEM), specifically Root Mean Squared Error Approximation (RMSEA). use simulations to show that naive implementations RMSEA have a downward bias presence and, thus, overestimate model goodness-of-fit. Unfortunately, many state-of-the-art software packages report biased form RMSEA. As consequence, scientific community been accepting much larger fraction models with non-acceptable fit. propose bias-correction for based information-theoretic considerations take into account expected misfit person fully observed data. The corrected is asymptotically independent proportion misspecified models. Importantly, results computation identical if there no