The paper discusses the analysis of variance (ANOVA) methods for gene expression microarray data, which are used to normalize data and estimate changes in gene expression while accounting for potential confounding effects. The authors demonstrate that ANOVA can effectively handle the inherent "noise" in microarray data, providing a framework for the general analysis and interpretation of such data. They use two designed experiments to compare the reproducibility of estimated changes in expression levels using ANOVA. The first experiment, a Latin square design, involves comparing human liver and muscle tissue samples using two arrays, while the second experiment uses a reference sample design with placenta as a reference. The results show that ANOVA methods can produce reliable estimates and confidence intervals for gene expression differences, even when the data are not normally distributed. The authors also compare the results from the two experiments, finding high agreement in the estimated differences in gene expression. The paper concludes by discussing the advantages of ANOVA methods over simple ratio-based approaches and the importance of proper experimental design in microarray studies.The paper discusses the analysis of variance (ANOVA) methods for gene expression microarray data, which are used to normalize data and estimate changes in gene expression while accounting for potential confounding effects. The authors demonstrate that ANOVA can effectively handle the inherent "noise" in microarray data, providing a framework for the general analysis and interpretation of such data. They use two designed experiments to compare the reproducibility of estimated changes in expression levels using ANOVA. The first experiment, a Latin square design, involves comparing human liver and muscle tissue samples using two arrays, while the second experiment uses a reference sample design with placenta as a reference. The results show that ANOVA methods can produce reliable estimates and confidence intervals for gene expression differences, even when the data are not normally distributed. The authors also compare the results from the two experiments, finding high agreement in the estimated differences in gene expression. The paper concludes by discussing the advantages of ANOVA methods over simple ratio-based approaches and the importance of proper experimental design in microarray studies.