Table 1.

Comparisons of t- and deviance tests in overdispersed logistic regression and log-linear models and a test based on a Bayesian model

	Group 1a		logistic regression		log-linear model		Bayesian model
	library 1	library 2	t-testc	deviance test	t-testc	deviance test	Ed
1b	0	0	0.645	0.115	0.003	0.001	0.01
2	2	2	0.485	0.122	0.002	0.002	0.02
3	5	5	0.383	0.133	0.003	0.005	0.04
4	10	10	0.324	0.149	0.007	0.01	0.05
5	20	20	0.291	0.183	0.02	0.025	0.07
6	50	50	0.324	0.29	0.104	0.117	0.11
7	100	100	0.494	0.508	0.376	0.404	0.12

^aTag counts in group 1 are artificially increased towards the levels observed in group 2 (which are held fixed). Tag counts in group 2 are 312, 549, 246, 65, 41, and 52. The library sizes and tag counts in group 2 are taken from Baggerly et al. [15].

^b The empirical tag counts 0.506, and 0.494 are used to replace the zero counts in group 1[15].

^c The t-test here is testing the hypothesis that β = 0.

^d E, the Bayes Error Rate, is listed. [26].