Table 3. Imputation accuracy for the two-step approach compared with direct imputation stratified per genotype class.

Minor allele frequency	N	Two stepa accuracy, mean (SD)	Direct accuracy, mean (SD)	P-valueb
Common homozygotes (major–major)
>0–1%	25 087	0.998 (0.005)	0.999 (0.004)	—
1–5%	25 087	0.998 (0.006)	0.998 (0.006)	—
5–10%	18 613	0.996 (0.009)	0.996 (0.01)	—
Heterozygotes (minor–major)
>0–1%	25 087	0.830 (0.237)	0.806 (0.244)	3.4e-6
1–5%	25 071	0.915 (0.145)	0.886 (0.172)	<1e-15
5–10%	18 613	0.955 (0.083)	0.942 (0.106)	<1e-15
Rare homozygotes (minor–minor)
>0–1%	0	—	—	—
1–5%	1248	0.895 (0.241)	0.826 (0.294)	<1e-15
5–10%	6746	0.925 (0.204)	0.888 (0.245)	<1e-15

^aFirst imputation filter on r²>0.3.

^bComparing groups, heterozygotes and rare homozygotes in each minor allele frequency bin.