Table 9.

Overview of Genotype Misclassification Models and Estimated Single Nucleotide Polymorphism Genotyping Error Size in the Literature

Error Model	Description	Parameterization	Reference No(s).	Estimated Error Size
“General” (model A)	No restrictions	${\text{π}}_{ij}$ = Prob(observed genotype = i|true genotype = j), i, j = 0, 1, 2	12, 22, 41
“Allelic dropout”	1 allele signal vanishes beneath white noise, and thus Prob(hom→het)a < Prob(het→hom)	${\text{π}}_{01}>{\text{π}}_{10}$ and ${\text{π}}_{21}>{\text{π}}_{12}$	18, 19
“Zero-corner” (model B)	Hom majorb never misclassified as hom minor and vice versa	${\text{π}}_{20}=0$, ${\text{π}}_{02}=0$	19
“Symmetrical” (model C)	Misclassification does not differ for hom major or hom minor	${\text{π}}_{10}={\text{π}}_{12}$, ${\text{π}}_{01}={\text{π}}_{21}$, and ${\text{π}}_{20}={\text{π}}_{02}$
“Hom-het”	Zero-corner and symmetry as described above	Prob(hom→het) = :ν, Prob(het→hom) = :μ	20, 24, 31
“Directed error”	Error described per allele	Prob(A→a) = :μ, Prob(a→A) = :νc	9, 11, 12, 23, 24, 31, 42
“Allele-independent” (also “stochastic error”) (model D)	Error described per allele, assumed independent from allele	Prob(A→a) = Prob(a→A): = ϵ (see also Table 3)	9, 12, 13, 29, 41, 43, 44	ϵ = 0.0074 (29); 1 study with 1,027 persons genotyped twice (30 discordances)
	Special case of symmetrical model; related to zero-corner model; if e is small, then e2 is close to 0
“Uniform error”	Uniform misclassification probabilities	${\text{π}}_{ij}$ = ϵ for i ≠ j, i, j = 0, 1, 2	7, 8, 10, 18, 21, 22, 32	ϵ = 0.015 (27); 1 study with 1,473 persons genotyped twice (2 discordances)

^aProb(hom→het): transition probability for the true homozygous genotype (either minor or major allele) being misclassified as heterozygous; Prob(het→hom), vice versa.

^bHom minor or hom major: homozygous for the minor or major allele, respectively.

^cProb(A→a): transition probability for the true major allele A being misclassified as the minor allele a; Prob(a→A), vice versa.