To utilize the data fully, statistical methods need to be developed on two fronts. Many obvious methodological refinements could increase yield and reduce error rates. Most importantly, even though this new method can be applied to tasks that were previously impossible, its power can be further enhanced by incorporating ideas behind existing methods. In our phasing examples, LRP on its own was performing very well, but for more difficult regions or with a smaller sample typed, local phasing, which in effect could generate more informative markers than bi-allelic ones, could assist in both phasing and the detection of IBD sharing of shorter regions. Apart from being the first procedure that can systematically impute haplotypes into completely untyped individuals, our method can also be used in conjunction with previous methods for the imputation of untyped variants into individuals with other markers typed21,22. In another front, we emphasize that proper, and sometimes sophisticated, statistical methods are needed in situations where valid estimates and measures of statistical significance are required.
