The present invention provides methods and computer programs for detecting variations in measurements of a biological variable, e.g., variations in cell viability, under different conditions, e.g., with or without treatment of a drug, under the treatments of different drugs, or under different environmental conditions. The methods and programs of the invention determine a metric of difference between measurements under different condition, and make use of predetermined errors of the measurements, e.g., errors determined from an error model, to estimate the errors in the metric of difference. Such an approach provides a more accurate estimate of the errors of measurements, which, in turn, provides a more accurate estimate of the error of the difference metric.