The widely used processing method for cDNA microarray data involves background correction, log-ratio transformation and data normalization before the statistical testing can be done. Here we propose a method that avoids the log-transformation step in view of its drawbacks, but goes directly to normalization after background correction. This method could better estimate the ‘noise’ effect by utilizing the information more effectively. Simulation studies were carried out to compare the feasibility and efficiency of this approach for eliminating experimental ‘noise’ with the log-ratio approach. Results showed that our approach worked well and the method was more robust and powerful than the log-ratio approach.
