Thinking outside the curve, part I: modeling birthweight distribution

Table 5 Confidence Interval Coverage Probabilities in Simulation Studies

C	Population Size	Bias adjustment included	Bias adjustment omitted
		Number & Percentage of Intervals Containing Targets (Mixture parameters)	Number & Percentage of Intervals Containing Targets (Mixture parameters)
2.0	200,000	106 (88.3)	50 (41.7)
	1,000,000	110 (91.7)	66 (55.0)
	Infinite	105 (87.5)	64 (53.3)
2.5	200,000	110 (91.7)	61 (50.8)
	1,000,000	112 (93.3)	75 (62.5)
	Infinite	114 (95.0)	74 (61.7)
3.0	200,000	110 (91.7)	67 (55.8)
	1,000,000	112 (93.3)	80 (66.7)
	Infinite	115 (95.8)	83 (69.2)
3.5	200,000	111 (92.5)	68 (56.7)
	1,000,000	115 (95.8)	83 (69.2)
	Infinite	118 (98.3)	88 (73.3)
4.0	200,000	112 (93.3)	69 (57.5)
	1,000,000	118 (98.3)	85 (70.8)
	Infinite	118 (98.3)	91 (75.8)
4.5	200,000	113 (94.2)	74 (61.7)
	1,000,000	118 (98.3)	87 (72.5)
	Infinite	118 (98.3)	96 (80.0)
5.0	200,000	116 (96.7)	78 (65.0)
	1,000,000	118 (98.3)	89 (74.2)
	Infinite	118 (98.3)	98 (81.7)

The row with "C" = 2 and "Population size" = 200,000 identifies the numbers and percentages of confidence intervals containing their targets of mixture parameters, based on 10 repetitions in each of which 25 samples of size 50000 were simulated from a 4-component normal mixture with 12 parameters; results under the heading of "Bias adjustment included" are based on Equation (7) with C = 2, results under the heading of "Bias adjustment omitted" are based on Equation (6) with C = 2, and the 25 samples of size 50000 had overlap consistent with a population size of 200,000. Other rows correspond to different choices of C and/or population sizes.

ISSN: 1471-2393