Table 7.

Percent Bias, N = 500, Simulation B

	Listwise		Log		CART		Prune		RF		MI
	30%	50%	30%	50%	30%	50%	30%	50%	30%	50%	30%	50%
${\overline{X}}_2$
Linear	−3.539	13.265	−.561	−.300	−1.206	1.504	−1.277	3.017	−.268	−2.303	−.602	.549
One split	−11.009	−11.891	−.066	1.011	−.228	−.056	−.196	−.148	4.392	3.425	−.466	−.723
Two splits	−2.615	−4.019	−.524	−.675	.466	−.450	.902	−.377	1.752	1.151	−.075	−.416
Three splits	−9.912	−12.207	−.781	−.836	−.067	−.204	−.019	−.686	1.328	1.498	−.429	−1.130
${\overline{Y}}_2$
Linear	−2.029	9.398	.117	.073	−.248	1.113	−.276	2.176	.431	−1.748	.050	.351
One split	−7.623	−8.479	.501	.878	.324	−.137	.365	−.212	3.427	2.248	−.291	−.371
Two splits	−1.598	−2.848	−.166	−.102	.411	−.291	.848	−.232	1.555	1.384	.188	−.226
Three splits	−6.520	−8.206	.027	.100	.393	.226	.421	−.108	1.322	1.821	.081	−.325
${\mathrm{\sigma }}_{X_2}^2$
Linear	−1.399	−1.511	−1.201	.163	−1.306	.025	−1.391	−.415	−.449	−.760	−1.694	−1.706
One split	−3.089	−1.447	−.135	.660	−.307	.227	−.488	−.097	.496	−.257	−1.070	−1.998
Two splits	−3.605	−4.119	−3.287	−2.269	.217	1.774	.100	1.877	.763	2.117	−.376	−1.016
Three splits	−2.389	−.929	−.341	.081	.710	−.298	.765	−.665	1.086	−.832	−.598	−1.057
${\mathrm{\sigma }}_{Y_2}^2$
Linear	−1.528	−4.661	−1.059	−.999	−.758	−2.315	−.572	−3.221	.904	−3.285	−2.132	−3.859
One split	−8.056	−4.375	2.089	1.924	.766	−1.429	.848	−1.383	.799	−2.195	−2.469	−3.490
Two splits	−10.246	−12.876	−9.938	−9.393	.827	.327	−.005	1.113	2.925	−1.016	−1.796	−3.497
Three splits	−6.431	−2.683	−1.089	.974	1.053	.362	1.470	−.183	2.164	.956	−1.693	−2.439
${\mathrm{\sigma }}_{X_2,Y_2}$
Linear	−1.257	−7.225	−.426	.079	.039	−1.873	.172	−3.549	2.414	−2.223	−2.283	−5.073
One split	−12.295	−6.004	2.036	3.234	.923	−1.500	.976	−1.685	1.942	−3.757	−2.415	−4.309
Two splits	−16.170	−19.752	−15.135	−13.396	1.429	1.317	−.051	2.142	3.808	.652	−1.479	−5.027
Three splits	−10.144	−5.970	−2.461	−.719	1.393	−1.654	1.905	−1.877	2.766	−2.675	−1.766	−4.876

Note. Log = logistic regression; CART = classification and regression trees; Prune = pruned CART analysis; RF = random forests; MI = multiple imputation.