. Author manuscript; available in PMC: 2009 Jul 15.

Published in final edited form as: Biostatistics. 2007 Jun 12;9(1):137–151. doi: 10.1093/biostatistics/kxm018

Table 3.

Mean estimates ofp and λ for models I and II over 100 simulations for N = 1000 observations simulated from “a mixture of 2 chi-square distributions, constant p”: $y_{01} \sim χ_{1}^{2}, y_{02} \sim χ_{1}^{2}, y_{11} \sim χ_{3}^{2}$ , and $y_{12} \sim χ_{3}^{2}$ . Empirical standard errors are given in parenthesis

Model	λ₁	λ₂	p = 0.5	Coverage^† for p	Log-likelihood
I, K= 0	0.20 (0.02)	0.20 (0.02)	0.51 (0.05)	0.94	−3152.49 (66.9)
I, K= 1	0.20 (0.02)	0.20 (0.02)	0.51 (0.05)	0.94	−3152.47 (66.9)
I, K= 2	0.19 (0.03)	0.18 (0.03)	0.48 (0.12)	0.74	−3138.60 (67.8)
II, K= 1			0.44 (0.06)	0.39	−3893.36 (56.6)
II, K= 2			0.44 (0.06)	0.38	−3862.35 (69.3)

Model	λ₁	λ₂	p = 0.5	Coverage^† for p	Log-likelihood

I, K= 0	0.31 (0.03)	0.30 (0.03)	0.04 (0.05)	0.77	−4061.50 (39.3)
I, K= 1	0.30 (0.03)	0.30 (0.03)	0.05 (0.07)	0.67	−4061.64 (40.3)
I, K= 2	0.28 (0.06)	0.27 (0.06)	0.08 (0.09)	0.66	−4054.07 (40.7)
II, K= 1			0.60 (0.05)	0.00	−4327.55 (40.5)
II, K= 2			0.59 (0.06)	0.00	−4253.37 (40.6)

^†

Coverage for likelihood-ratio-based confidence intervals.