Table 1.

Storage size required for various compressed reference haplotype formats

Dataset	Size (MB) for format:
	vcf.gz	jlso	bref3	m3vcf.gz	imp5
sim 10K	49	6	18	7	57
sim 100K	472	56	81	29	565
sim 1M	4660	776	415	NA	5650
1000G chr10	459	188	342	116	621
1000G chr20	200	89	154	52	279
HRC chr 10	3346	1328	1156	529	3636
HRC chr 20	1510	706	554	253	1610

Note: Here, vcf.gz is the standard compressed VCF format; jlso is used by MendelImpute; bref3 is used by Beagle 5.1; m3vcf.gz is used by Minimac 4; and imp5 is used by Impute 5. For all jlso files, we chose the maximum number of unique haplotypes per window to be $d_{\text{max}}=1000$. Note we could not generate the m3vcf.gz file for the sim 1 M panel because it required too much memory (RAM).