Minimizing the conditional variance based on the A matrix (MCA ) makes sense since the aim is to predict the missing genotypes and thus, conditional on (dense) genotypic information, to minimize the unexplained genetic variation.
The multiple-imputations method alleviates the issue of missing genotype data by imputing all missing genotypes numerous times conditional on observed genotype data at neighboring markers and storing these imputations.
They also used an expression for the conditional variance of the missing genotypes that is valid only for multivariate normal variables.
In contrast, in LDMIP, conditional probabilities of the missing genotypes, given all observed genotypes at locus j and linked loci, that are computed approximately by iterative peeling and combine both linkage and LD information, are used to impute genotypes.
Then, conditional on the inferred haplotypes, all missing genotypes were imputed using Beagle's hidden Markov model.
A fast and flexible statistical model for large-scale population genotype data: applications to inferring missing genotypes and haplotypic phase.
Scheet, P. & Stephens, M. A fast and flexible statistical model for large-scale population genotype data: applications to inferring missing genotypes and haplotypic phase.
The mutual intersection of all data sets yielded 19,372 diallelic, autosomal SNPs with experimentally determined genotypes (i.e., no imputation of missing genotypes was performed).
In addition, 39,033 SNPs were excluded owing to low genotyping (with > 5% missing genotypes per marker) and 198,553 SNPs, owing to minor allele frequency of < 1%.
The following filters were used: MAF of 0.05, Hardy-weinberg p = 0, exclude markers with >0.5 missing genotypes and less than 0.5 nonzero genotypes.
