Table 2.

Simulations based on empirical data: a comparison of type I errors and power for various methods under 2 disease model in 500 simulations. Standard errors are given in parentheses. For the first model, we generate data for 800 cases and 800 controls with ORs 1.5, 1.75, and 2; for the second model, we generate data for 400 cases and 400 controls with ORs 1.25, 1.5, and 1.75

	Method	Type I error	Power
			OR = 1.5	OR = 1.75	OR = 2
Model 1†	Single-locus scan	0.048 (0.010)	0.318 (0.015)	0.648 (0.015)	0.914 (0.013)
Phase known	Full haplotype	0.052 (0.010)	0.284 (0.020)	0.590 (0.022)	0.870 (0.015)
	CLADHC	0.056 (0.010)	0.256 (0.020)	0.546 (0.022)	0.854 (0.016)
	SHARE	0.050 (0.010)	0.336 (0.021)	0.654 (0.021)	0.928 (0.012)
Phase unknown	Haplotype score	0.035 (0.008)	0.288 (0.020)	0.544 (0.022)	0.863 (0.015)
	SHARE	0.054 (0.010)	0.326 (0.021)	0.650 (0.021)	0.900 (0.013)
			OR = 1.25	OR = 1.5	OR = 1.75
Model 2‡	Single-locus scan	0.046 (0.007)	0.176 (0.017)	0.608 (0.015)	0.916 (0.012)
Phase known	Full haplotype	0.046 (0.009)	0.184 (0.017)	0.616 (0.022)	0.920 (0.012)
	CLADHC	0.062 (0.011)	0.138 (0.015)	0.548 (0.022)	0.882 (0.014)
	SHARE	0.046 (0.009)	0.182 (0.017)	0.678 (0.021)	0.952 (0.010)
Phase unknown	Haplotype score	0.050 (0.010)	0.158 (0.016)	0.586 (0.022)	0.900 (0.013)
	SHARE	0.044 (0.009)	0.190 (0.018)	0.666 (0.021)	0.942 (0.010)

^†The unscored disease-causing locus is best captured by haplotypes based on 2 tagSNPs.

^‡Two tagSNPs separated apart carry disease risk additively.