TABLE 1.

Comparison of accuracy of BLIMP and IMPUTE for frequency and individual-level genotype imputations. The RMSE and Error rate, defined in the text, provide different metrics for assessing accuracy; in all cases BLIMP was very slightly less accurate than IMPUTE. The “naive method” refers to the strategy of estimating the sample frequency of each untyped SNP by its observed frequency in the panel; this ignores information in the observed sample data, and provides a baseline level of accuracy against which the other methods can be compared.

Frequency imputation
	RMSE	Error Rate
naive method	0.0397	NA
BLIMP (phased panel)	0.0157	NA
BLIMP (unphased panel)	0.0159	NA
IMPUTE	0.0154	NA
Individual genotype imputation
	RMSE	Error Rate
BLIMP (phased panel)	0.2339	6.46%
BLIMP (unphased panel)	0.2407	6.77%
IMPUTE	0.2303	6.30%