Table 4.

Bias (AIC) in estimating the relative expected Kullback–Leibler divergence (rEKL) values for 0% and 50% groups of MSAs for 3 simulations. Bias(AIC) = rEKL − E(AIC). The Bias(AIC) were the same for the GHOST mixture models across all groups of MSAs, so only one value is shown for the mixture models. The Bias(AIC) for the partition models were very similar to each other, so only the Bias(AIC) from 0% and 50% groups are shown in the table

Simulation	Group	Mean bias (AIC)
		P-LEL	P-UEL	M-UGP	M-LGP
Standard	$0\mathrm{\%}$	15.4	11.4	12.4	16.8
	$50\mathrm{\%}$	20.5	20.0
Mild	$0\mathrm{\%}$	−40.1	−77.8	−86.0	−21.2
	$50\mathrm{\%}$	−35.5	−61.2
Extreme	$0\mathrm{\%}$	−46.8	−101.8	−91.7	−14.3
	$50\mathrm{\%}$	−40.4	−87.0