When the distributional assumption is severely violated, some distribution-free imputation methods may be considered.
These results implied that also pedigree-free imputation software (Minimac and IMPUTE2) yielded accurate genotypes for whole sequence variants in this data set.
IMPUTE2 is a hidden-Markov model based pedigree-free imputation approach with methodological similarities to Beagle, MaCH, and fastPHASE, which have been compared to each other in several studies [ 4, 19, 20].
Thus, the advantage of whole-genome sequence data over very high-density genotyping is still unclear, even with total absence of sequencing errors and of missing data (which is equivalent to an error-free imputation process).
It is likely that RFI and kNNI produced comparable levels of accuracy because both use a similar model free approach for imputation that involves weighting a selected set of k important variables according to a distance metric (Lin and Jeon 2006).
In practice, sequencing data analysis and imputation are not free of charge in considering compute-intensive analysis.
The method is also model-free, manages missing values without imputation techniques (i.e., pair-wise feature correlations can be estimated by excluding only samples with missing values), robust to data noises and provides effective visualization tools.
In contrast, multiple imputation was virtually bias-free and gave robust estimates in all the scenarios.
All statistical analyses were carried out using SAS® software (version 9.2, SAS Institute Inc., Cary, NC) and R.2.12.0 software (free) using the Design, SurvIDINRI (for NRIs) and Multivariate Imputation by Chained Equations packages (http://www.multiple-imputation.com/).
MGB-2: human Mammaglobin B; EOC: epithelial ovarian cancer; HOSE: human ovarian surface epithelium; qRT-PCR: quantitative Reverse Transcription-Polymerase Chain Reaction; IHC: immunohistochemistry; H&E: hematoxilin and eosin; OS: overall survival; DFS: disease-free survival; PFS: progression free survival; HR: hazard ratio; CI: confidence interval; OR: odds ratio; MI: multiple imputation.
Multiple imputation was used in dealing with missing body composition (fat mass, and fat free soft tissue mass) and metabolic outcome variables (fasting blood glucose, fasting insulin, HOMA, HDL, LDL, triglycerides, and total cholesterol).
