A three-source capture-recapture estimate of the number of new HIV diagnoses in children in France from 2003–2006 with multiple imputation of a variable of heterogeneous catchability

Table 1 Preliminary log-linear analyses assuming homogeneity of capture within each source

Log-linear models	$\hat{n}$	$\hat{N}$	95% CI	df	G²	p	AIC	DIC
Dependencies between sources
LaboVIHDOVIH, LaboVIHEPF, DOVIH*EPF	126	342	259,573	0	0	1	0	0
LaboVIHDOVIH, LaboVIHEPF	23	239	225,263	1	18.83	<10^-4	16.83	16.89
LaboVIHDOVIH, DOVIHEPF	58	274	243,331	1	3.78	0.05	1.78	1.84
LaboVIHEPF, DOVIHEPF	153	369	294,521	1	0.24	0.63	−1.76	−1.71
LaboVIH*DOVIH, EPF	29	249	234,272	2	18.49	<10^-4	14.49	14.6
LaboVIH*EPF, DOVIH	51	267	245,300	2	30.12	<10^-4	26.12	26.23
DOVIH*EPF, LaboVIH	85	301	268,349	2	5.96	0.05	1.96	2.07
LaboVIH, DOVIH, EPF	49	265	246,292	3	30.2	<10^-4	24.20	24.36

DOVIH: Mandatory HIV case reporting; EPF: ANRS French Perinatal Cohort; LaboVIH: Laboratory surveillance of HIV testing activity.
$\hat{n}$ : Estimate of the number of diagnoses not reported to any source; $\hat{N}$ : Estimate of the number of diagnoses; 95% CI: 95% confidence interval for $\hat{N}$ _; df: number of degrees of freedom; G²: deviance statistic; p: p-value of the deviance goodness-of-fit test; AIC: Akaike Information Criterion; DIC: Draper Information Criterion.

ISSN: 1471-2334