Generally speaking, in order to determine the best fitting model, researchers must
take into account several criteria (for a thorough discussion on model fit, see Bond &
Fox, 2001). These criteria include: (1) statistical criteria, for example, deviance
statistics, information criteria, and misfitting items (discussed in more detail in the
section ‘Item quality and item redundancy’), and (2) non-statistical criteria, for
example, the underlying theoretical basis of the instrument, and aspects of instrument
content. Collectively, researchers must weigh how these criteria inform interpretations
of model fit.