Domain estimation under informative linkage

Figures & data

Figure 1. Simulation A with fixed domain effects: Relative bias (%) of domain mean estimators. Horizontal axis represents the different domains.

Figure 2. Simulation A with random domain effects: Relative bias (%) of domain mean estimators. Horizontal axis represents the different domains.

Figure 3. Simulation A with fixed domain effects: Relative RMSE (%) of domain mean estimators. Horizontal axis represents the different domains.

Figure 4. Simulation A with random domain effects: Relative RMSE (%) of domain mean estimators. Horizontal axis represents the different domains.

Figure 5. Simulation A with fixed domain effects: Coverage (nominal = 95%) of domain mean estimators. Horizontal axis represents the different domains.

Figure 6. Simulation A with random domain effects: Coverage (nominal = 95%) of domain mean estimators. Horizontal axis represents the different domains.

Figure 7. Simulation B with random domain effects: Relative Bias (%) of domain mean estimators. Horizontal axis represents the different domains.

Figure 8. Simulation B with random domain effects: Relative RMSE (%) of domain mean estimators. Horizontal axis represents the different domains.

Figure 9. Simulation B with random domain effects: Coverage (nominal = 95%) of domain mean estimators. Horizontal axis represents the different domains.

Table 1. Summary measures of quality of record linkage for linked Brazilian data.

	Comparison weights (FS) linkage		Classification tree (Bagging) linkage
Averages of linkage quality metrics	Error level 1	Error level 2	Error level 1	Error level 2
Average counts for all pairwise comparisons
True links made	1220	807	1250	1211
False links ignored	634,114	632,701	633,105	634,664
True links ignored	1334	1748	1308	1354
False links made	60	473	30	69

Table 2. Average probabilities of correct linkage by linkage method and level of error (Block = State) for linked Brazilian data.

	Comparison weights (FS) linkage		Classification tree (Bagging) linkage
Probability of correct linkage	Error level 1	Error level 2	Error level 1	Error level 2
Block 1	0.95	0.74	0.97	0.95
Block 2	0.95	0.52	0.97	0.93
Block 3	1.00	0.92	1.00	1.00
Block 4	0.95	0.72	0.98	0.97

Table 3. Summary statistics for industry estimates of production using sample-to-register linkage and selected estimators for Brazilian linked data.

	Comparison weights (FS) linkage						Classification tree (Bagging) linkage
	Error level 1			Error level 2			Error level 1			Error level 2
Industry	WT	BL	MLE	WT	BL	MLE	WT	BL	MLE	WT	BL	MLE
	Relative Bias (%)
Crops	−0.04	0.23	0.14	−1.69	0.64	0.33	−0.10	0.03	0.14	−0.18	0.14	0.04
Livestock	0.13	0.26	0.14	−0.83	0.51	0.21	−0.21	−0.15	−0.12	0.05	0.22	0.02
Forestry	−0.05	0.16	−0.14	−0.73	0.32	−0.27	−0.04	0.06	0.00	−0.06	0.13	−0.03
Fishery	0.25	−0.35	−0.36	5.23	−0.30	−0.67	0.11	−0.21	−0.05	0.61	−0.13	−0.09
	Relative RMSE (%)
Crops	1.66	1.74	1.39	2.51	2.78	2.38	1.58	1.59	1.32	1.84	1.91	1.51
Livestock	2.14	2.24	1.52	2.25	3.20	1.89	2.21	2.24	1.46	2.17	2.25	1.45
Forestry	1.30	1.35	1.27	2.00	2.40	1.43	1.46	1.50	1.16	1.42	1.47	1.24
Fishery	1.75	1.84	1.75	5.81	3.96	2.74	1.63	1.67	1.52	2.03	2.06	1.76
	Coverage (%) of nominal 95% Gaussian confidence intervals
Crops	98.0	96.0	96.0	88.4	77.4	94.5	99.0	97.5	98.0	94.0	92.0	95.5
Livestock	86.0	85.0	92.0	93.2	67.8	90.4	90.5	87.5	94.0	88.0	86.0	92.0
Forestry	100.0	100.0	97.0	100.0	84.9	97.3	100.0	100.0	98.5	100.0	100.0	97.5
Fishery	99.5	100.0	98.5	54.8	91.1	96.6	99.5	100.0	97.5	99.0	98.0	97.0