Figure 2
From: Use of normalization methods for analysis of microarrays containing a high degree of gene effects
Comparison of normalization accuracy and precision using seven different analytical methods. (A) Mean relative error in normalization (scaling) factors determined by the seven different normalization methods using Dnl dataset. The relative error for each feature of the microarray is defined as |(estimated – truth)/truth|, where the mean relative error is derived by averaging all features in an array. Errors that are greater than 50% are not shown. (B) Coefficient of variation (CV) between D1 and gene-effect matched normalized Dnl data. D1 and Dnl are defined in Fig. 1. The x-axis in (A-B) corresponds to the total percentage of gene effects (shown as DEG) in a microarray, and up, down, and up/down indicates these effects as up-, down-, and an equal mixture of both, respectively. Xcorr: cross-correlation. Median: global median normalization.