Table 2.

Log-linear analyses incorporating variables of potential heterogeneous catchability

Log-linear models	$\widehat{\mathbf{n}}$	$\widehat{\mathbf{N}}$	95% CI ($\widehat{\mathbf{N}}$)		df	G2	p	AIC	DIC
Model 1: DO*EPF, LABO*EPF, EPF*place	154	370	270	469	100	145,75	0,00	−54,25	−142,31
Model 2: DO*EPF, LABO*EPF, EPF*region	154	370	270	469	100	133,74	0,01	−66,26	−154,32
Model 3: DO*EPF, LABO*EPF, EPF*année, Labo*année	154	370	270	469	95	141,78	0,00	−48,22	−131,88
Model 4: DO*EPF, LABO*EPF, EPF*place, EPF*region, DO*region	147	363	266	460	97	125,12	0,03	−68,88	−154,30
Model 5: DO*EPF, LABO*EPF, EPF*place, EPF*year	131	347	260	434	94	132,81	0,00	−55,19	−137,96
Model 6: DO*EPF, LABO*EPF, EPF*region, DO*region, EPF*year	141	357	261	452	96	121,31	0,02	−64,69	−146,59
Model 7: DO*EPF, LABO*EPF, DO*place, EPF*place, DO*region, Labo*region, EPF*region, EPF*year	171	387	271	503	93	112,72	0,07	−73,28	−155,17
Model 8: DO*EPF, LABO*EPF, DO*place, EPF*place, DO*region, Labo*region, EPF*region	171	387	271	503	96	112,48	0,05	−73,52	−158,06

DO: Mandatory HIV case reporting; EPF: ANRS French Perinatal Cohort; Labo: Laboratory surveillance of HIV testing activity.

$\widehat{n}$: Estimate of the number of diagnoses not reported to any source; $\widehat{N}$: Estimate of the number of diagnoses; 95% CI: 95% confidence interval for $\widehat{N}$_; 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

Place: place of birth; region: region of diagnosis; year: year of diagnosis.