A caveat is that the number of loci required can be quite large if the fraction of missing data is large or the number of taxa large (when p << 1 and or q << 1, the lower bound on number of loci k scales as 1/(p q) (expanding the expression Theorem 2 part (ii) in a Taylor series approximation)).
Performance improvements of likelihood-based methods over Rg become more significant when identification is performed on the dataset with higher percentage of missing data and larger noise.
Markers are removed if the percentage of missing data was larger than 5% or if they are not in Hardy-Weinberg equilibrium (p > 0.0001 for control group).
Generally, problems with missing data in large observational databases have received a great deal of attention in the literature, and multiple imputation (MI) is the primary technique for handling missing data.
Missing data were estimated using Beagle v3.3, which enables the inference of haplotypes and imputation of sporadic missing data in large-scale genotype datasets (Courtois et al. 2013).
Though it often produces reasonable results, CCkNNI is severely limited when the amount of missing data is large (and hence the number of complete cases is small).
Furthermore, according to Philippe et al. (2004), missing data in large alignments do not significantly affect inferred phylogeny when resolving relationships among eukaryotic groups.
However, if the fraction of missing data is large, say in the order of 30%to50%0%, imputation methods must be applied with great caution (White et al., 2011).
Imputation of missing data in large regions of satellite imagery is necessary when the acquired image has been damaged by shadows due to clouds, or information gaps produced by sensor failure.
In the example considered in this paper, both the number of raters and the proportion of missing data were large, and therefore to find a model that could be fitted within the limits of computational feasibility was an important consideration.
