Current Mathematical Models for Analyzing Anti-Malarial Antibody Data with an Eye to Malaria Elimination and Eradication

Table 1

Gaussian mixture modelling analyses for determining seropositivity to AMA1 titre data in a sample of around 6400 individuals from Bioko Island using 90% as the cut-off value for the correct classification probability.

Number of components	AIC	Mean (SD)	Definition of	Cut-off values		Classification probabilities
			and
2	84601.2	59.3 (48.4)	,	95.9	202.9	31.2	12.7	56.1
		668.1 (450.4)
3	83395.2	35.8 (26.8)	,	44.6	109.8	19.3	13.8	66.8
		214.0 (115.4)	,	103.8	515.2	32.3	33.3	34.4
		848.3 (425.6)
4	82887.4	14.1 (9.1)	= 1, = 2, 3, 4	NA	37.2	—	17.0	83.0
		64.7 (32.4)	,	34.0	149.5	16.1	22.1	61.9
		252.2 (120.6)	,	135.4	560.3	36.4	31.3	32.3
		873.2 (420.6)

The best model is the one providing the lowest estimated value.
Mean and standard deviation (SD) are for each Gaussian component in the model ordered by the corresponding average titres.
and are the cut-off values for determining the seronegative and seropositive populations, respectively.
, , and are the estimated classification probabilities of seronegative, indeterminate, and seropositive individuals, respectively.