Table 3.

The Pearson's correlation coefficient r between log-adjusted read counts and log-adjusted TaqMan values

Method	Correlation
Unadjusted	0.6650**
7mer	0.6680**
GLM	0.6874**
MART	0.6998*
BN	0.7086

We estimated the statistical significance of the improvement in correlation using the BN method over the other methods using a simple boostrap procedure. A bootstrap sample is formed by sampling, with replacement, 648 genes from the original set of the same size. The correlation is then computed for this set, using the adjusted count from each method. We repeated this procedure one million times, and counted the number of times each of the competing methods achieved a higher correlation than the BN method. Those marked with a single asterisk achieved a higher correlation fewer than 1000 times, resulting in a P < 10⁻³. Those marked with two asterisks achieved a higher correlation in none of the bootstrap samples, indicating a P < 10⁻⁶.