imp {dotgen}

Impute missing genotypes

Description

Impute missing genotype calls with values inferred from non-missing ones.

Usage

imp_avg(g, ...)

imp_cnd(g, ...)

Arguments

g

genotype matrix, one row per sample, and one column per variant.

...

additional parameters.

Details

A seemingly naive way to impute a missing value is to use the average of all non-missing values per variant, imp_avg(). Besides simplicity, this imputation by average has the advantage of approximating the correlation among test statistics (i.e., Z-scores) when the original association analyses were performed with missing values unfilled, which is a common practice. This naive approach is the defualt for the correlation calculator cst().

An advanced imputation approach is based on the conditional expectation method, imp_cnd(), that explores the relationship between variants and borrows information from variants other than the target one when making guesses. The sample correlation among variants imputed this way is closer to the true LD, and may improve power. However, after this imupation one must re-run the association analyses with imputed variants to avoid inflation in Type I error rates.

Value

imputed genotype matrix without any missing values.

Functions

• imp_avg: imputation by average.

• imp_cnd: imputation by conditional expectation