The paper presents a statistical approach to assess gene significance from cDNA microarray expression data, aiming to control the percentage of false positives while improving the detection of false negatives. The method uses two interconnected mixed linear models to account for variability across and within genes, allowing for the analysis of various experimental designs and multiple biological samples. The models are flexible and can be used to evaluate the statistical power of experimental designs, helping researchers select appropriate replicate numbers. The approach is illustrated through analyses of published experiments on human cancer and yeast cells, demonstrating its effectiveness in identifying significant gene expression differences. The results highlight the importance of proper statistical methods in distinguishing biologically important changes from random variation.The paper presents a statistical approach to assess gene significance from cDNA microarray expression data, aiming to control the percentage of false positives while improving the detection of false negatives. The method uses two interconnected mixed linear models to account for variability across and within genes, allowing for the analysis of various experimental designs and multiple biological samples. The models are flexible and can be used to evaluate the statistical power of experimental designs, helping researchers select appropriate replicate numbers. The approach is illustrated through analyses of published experiments on human cancer and yeast cells, demonstrating its effectiveness in identifying significant gene expression differences. The results highlight the importance of proper statistical methods in distinguishing biologically important changes from random variation.