Model fit decisions largely depend on the specifications of particular model fit indices (AIC vs. BIC vs. CAIC) leading to conflicting interpretations. Instead of making model fit decisions based on abstract criteria we pursue an approach based on utility for stakeholders with a generalized sub-dimensional (GSM) model. Examining results from a large data set collected from the Performance Assessment for California Teachers (PACT) instrument, we employ traditional model fit analyses and compare them to the GSM approach. Results from this validation study add more information on the appropriate uses and interpretations of PACT score data, particularly on issues related to sub-scale reliabilities.