Sero-epidemiology of Peste des petits ruminants virus infection in Turkana County, Kenya

BMC Veterinary Research

Table 3 Mixed model analyses, variance and summary intra-correlation coefficient (ρ) for exposure to PPRV infection in sheep and goat data

	Sheep n = 431		Goats n = 538
Variable	Odds ratio [95% CI]	LRT ^¥ P-value	Odds ratio [95% CI]	LRT ^¥ P-value
Sex	-			0.0023
Male			1
Female			0.13 [0.04, 0.5]
Age				0.000
Young	1	0.000	1
Middle age	0.2 [0.07, 0.35]		0.04 [0.02, 0.12]
Adult	0.6 [0.3, 1.18]		0.1 [0.02, 0.66]
Vaccination status			-
No	1	0.0004
Yes	4.5 [1.94, 10.6]
Administrative division				0.0005
Kaaleng	1	0.0036	1
Kakuma	1.1 [0.27, 4.27]		0.7 [0.27, 1.70]
Kibish	4.6 [1.25, 16.70]		3.6 [1.39, 9.53]
Loima	3.1 [0.79, 11.96]		1.2 [0.44, 3.21]
Lokichogio	3.3 [0.86, 12.64]		1.7 [0.67, 4.52]
Oropoi	11.7 [2.36, 57.70]		6.8 [2.29, 20.34]
*Agesex interaction**	-		2.8 [1.63, 4.88]	0.0002
Random effect –sublocation variance	0.61 [0.36, 1.04]		0.44 [0.18, 1.1]

LRT^¥: Likelihood ratio test.
*denotes age and sex interaction.
Random effect –sublocation: Sheep, likelihood ratio test versus standard logistic regression: chibar2(01) = 10.86; Prob> = chibar2 = 0.0005; ρ = 0.16; Goats, likelihood ratio test versus standard logistic regression: chibar2(01) = 2.07; Prob > =chibar2 = 0.075; ρ = 0.12. “chibar2(01)” test statistic tests whether random effects are greater than zero. The results of this likelihood ratio test shows that inclusion of sub-location random effect provided a substantially better fit than the multivariable logistic regression in Table 2 (at both 0.05 and 0.1 levels of significance (sheep data) and at 0.1 level of significance (goat data).

ISSN: 1746-6148