what does deviance measure
lack of fit
how can deviance be reduced to 0
making the model more complex and estimating values at every data point
what is wrong about making models more complex
it involves adding more predictors, makes the model less interpretable
what does information criterion account for
the fitting ability of the model
the complexity of the model
information criterion =
deviance + penalty for model complexity
what is used to assess deviance
AIC and BIC