Table III.

Proportional mean squared error of estimation for each parameter, from each model, with N_pairs = 1000

Parameter	Normal	Lognormal	Ordinal	LGP	NegBin
Mean	0.0026	0.0007	--	0.0024	0.0024
Phenotypic Variance	0.0267	0.0014	--	0.0169	0.0683
MZ Covariance	0.0624	0.0062	--	0.0185	0.0710
DZ Covariance	0.0936	0.0115	--	0.0214	0.0707
MZ Correlation	0.0220	0.0040	0.0025	0.0012	0.0012
DZ Correlation	0.0441	0.0088	0.0128	0.0033	0.0032
Additive-Genetic Variance	0.9931	0.1554	--	0.0733	0.1071
Shared-Environmental Variance	0.5316	0.0939	--	0.0506	0.0831
Nonshared-Environmental Variance	0.0868	0.0113	--	0.0205	0.0670
Heritability	0.9045	0.1548	0.1883	0.0577	0.0561
Shared-Environmentality	0.4363	0.0903	0.1138	0.0305	0.0294

notes: Proportional mean squared error is the ratio of the mean squared error of the estimator to the square of the true parameter value. Table presents data from the simulation condition in which samples consisted of 1000 twin pairs. Because of constraints necessary to identify the ordinal model, some of its parameters are not freely estimated or are redundant with other parameters. Generalized estimating equations analyses were conducted for each parameter. The pMSE for the MZ correlation did not significantly differ between the negative binomial and LGP models, but all other pairwise comparisons were statistically significant (smallest χ²(1df) = 11.55).