Model Fit, Comparison, and Checking

Model assessment involves both choice between competing models in terms of best fit, and checks to ensure model adequacy. For example, even if one model has superior fit, it still needs to be established whether predictions from the model check with, namely, reproduce satisfactorily, the observed data. Checking may also seek to establish whether model assumptions (e.g., normality of random effects) are justified, and whether particular observations are poorly fit (Berkhof et al., 2000; Sinharay and Stern, 2003).