Skip to main content

Table 1 Comparison of model fits to a single group of biological replicates

From: Statistical analysis and significance testing of serial analysis of gene expression data using a Poisson mixture model

     mean AIC mean BIC
  N tags k Poisson Negbin Mixture Poisson Negbin Mixture
BRAIN          
astrocytoma 14 1141 2.6 238.2 105.2 103.6 238.9 106.5 106.3
ependymoma 10 1205 2.3 152.4 80.9 75.0 152.7 81.5 76.1
glioblastoma 7 1197 2.3 139.5 57.6 53.0 139.4 57.5 52.8
medulloblastoma 18 1045 2.7 280.6 128.7 128.7 281.5 130.5 132.6
normal 8 1099 2.4 156.8 68.0 59.8 156.8 68.2 60.1
AML          
inv(16) 5 900 1.7 68.9 39.3 37.6 68.5 38.5 36.7
t(8;21) 5 1037 1.3 52.3 34.1 33.5 51.9 33.3 32.9
t(15;17) 5 709 1.8 127.7 46.0 38.9 127.3 45.2 37.9
t(9;11) de novo 4 954 1.8 58.5 34.6 30.1 57.9 33.4 28.5
t(9;11) treatment 3 1061 1.5 42.9 32.0 20.9 42.0 30.2 19.1
BREAST          
normal 6 1259 1.8 71.6 43.9 41.6 71.4 43.5 41.0
DCIS 4 598 1.3 25.5 24.0 21.0 24.9 22.8 20.1
invasive 3 1069 2.0 60.4 27.8 22.8 59.5 26.0 20.1
SKIN          
normal 4 1015 1.6 33.8 24.6 22.2 33.2 23.4 20.9
melanoma 3 992 1.8 38.2 24.0 19.6 37.3 22.2 17.4
  1. SAGE tag counts from fifteen sets of biological replicates were fit to log-linear (Poisson), negative binomial (overdispersed log-linear), and Poisson mixture models. The table contains the number of replicates (N), tags tested, and mean number of mixture components (k). For each model, mean goodness of fit scores calculated using Akaike's Information Criterion (AIC) and Bayesian Information Criterion (BIC) are shown. For both scores, a lower value indicates a better fit.