A new method for identifying reliably correlated variables

Post-genomic data analysis poses a new problem for identifying significantly reliably co-expressed genes. Because the gene co-expression can be generally measured by Pearson correlation (PC), in this paper, we formulate the problem as a new PC significance problem and present an analytical method for solving it via the Fisher transformation of PC. Traditionally, PC between two variables is tested to be equal to zero or not. In contrast, the proposed PC significance problem is to test if the PC is larger than a preset cutoff between 0 and 1, which thus extends the traditional one. We evaluate the new PC statistical inference on simulation data sets and show its goodness and power in PC statistical inference.