Table 4.

Summary of bivariate models fitted with three-way separable and 2DIMVAR1 residual models, spatial autocorrelation for column (${\varphi }_c$) and row (${\varphi }_r$) directions, residual log-likelihood (LL), and p-values of REML log-likelihood ratio test (REMLRT)

	Separable			2DIMVAR1
				BLeme		BLmat
Dataset	${\varphi }_c$	${\varphi }_r$	LL	${\varphi }_c$	${\varphi }_r$	${\varphi }_c$	${\varphi }_r$	LL	REMLRT
HS21	0.141	0.369	$-$2782	0.140	0.505	0.149	0.540	$-$2776	0.002 **
YK21	0.249	0.378	$-$2565	0.350	0.312	0.227	0.381	$-$2563	0.084 n.s
WN22	0.010	0.342	$-$1277	0.098	0.472	$-$0.052	0.183	$-$1274	0.064 n.s
HS22	0.193	0.127	$-$1245	0.222	0.116	0.216	$-$0.185	$-$1242	0.039 *
RS22	0.249	0.217	$-$195	0.434	0.263	0.256	0.213	$-$192	0.031 *
YK22	$-$0.067	0.061	$-$1040	$-$0.061	$-$0.027	$-$0.061	$-$0.037	$-$1039	0.307 n.s
WN23	0.029	$-$0.030	$-$312	0.135	$-$0.171	0.002	0.024	$-$306	0.002 **
HS23	0.238	0.155	$-$299	0.180	0.123	0.072	0.054	$-$298	0.240 n.s
RS23	$-$0.081	0.740	$-$2063	$-$0.151	0.839	$-$0.077	0.695	$-$2061	0.041 *
YK23	0.051	0.087	$-$2164	0.178	0.098	$-$0.077	0.086	$-$2162	0.104 n.s