Table 2.

Summary of mean absolute error (MAE) and continuous ranked probability score (CRPS) of the test set prediction with various average sample sizes for the pseudo site. All numbers are in percentages. The number of data years and the average number of sites of the training set are shown in parentheses. Sample size 0 corresponds to the case with no auxiliary data. The smallest MAE and CRPS of each area are in bold.

	Sample Size	MAE	CRPS	MAE	CRPS
Angola		Rural (1 year, 10 sites)		Urban (3 years, 25 sites)
	0	1.144	0.711	1.097	0.813
	4508 (GLMM)	1.058	0.677	1.099	0.814
	10	1.058	0.679	1.101	0.815
	100	1.067	0.682	1.094	0.811
	1000	1.047	0.675	1.101	0.813
		Rural (2 years, 4.5 sites)		Urban (5 years, 8 sites)
Liberia	0	2.866	2.228	1.559	1.033
	560 (GLMM)	2.379	1.919	1.589	1.053
	10	2.466	1.948	1.535	1.022
	100	2.418	1.931	1.547	1.025
	1000	2.377	1.915	1.561	1.037
		Rural (16 years, 7.9 sites)		Urban (17 years, 20.7 sites)
Ghana	0	0.753	0.507	0.925	0.647
	5653 (GLMM)	0.748	0.504	0.924	0.648
	10	0.741	0.499	0.928	0.650
	100	0.746	0.503	0.924	0.647
	1000	0.750	0.504	0.930	0.651
		Hhohho (6 years, 2 sites)		Manzini (6 years, 2 sites)
Swaziland	0	4.565	3.665	6.019	3.991
	1272 (GLMM)	4.452	3.530	6.157	4.100
	10	4.278	3.389	6.657	4.680
	100	4.420	3.483	6.427	4.392
	1000	4.451	3.530	6.273	4.202
		Shiselweni (6 years, 2 sites)		Lubombo (6 years, 2.2 sites)
	0	6.066	4.812	4.851	3.555
	1272 (GLMM)	6.117	4.982	4.610	3.328
	10	6.151	5.076	4.490	3.179
	100	6.134	5.040	4.506	3.194
	1000	6.130	5.012	4.601	3.314