Analysis of data from experiments using double labeling.

Frequently as a result of experiments in which two isotopes are used one is left with a sequence of samples, the ratio of labeling in each sample, and the problem of analyzing the ratios. Suppose that the experiments are designed so that one expects uniform labeling except for one or two special groups of samples. The problem, then, is to find these groups. Because of the variability in the count rate from sample to sample, the variance of the ratios differs from sample to sample making statistical analysis difficult. Furthermore, there is significant serial correlation in the sample disintegrations per minute for each of the isotopes. We have found that the serial correlation in the labeling ratio is small and of questionable significance in controls but becomes significant when there is a subsequence of samples in which the labeling ratio differs from that in the remainder of the gel. We examine the analysis of variance as a test for significant deviations in the labeling ratio and suggest a method for plotting deviations of labeling ratio from the average background labeling ratio. Finally, we develop a method of estimating the mean labeling ratio from the regression of disintegrations per minute of one isotope on those of the other isotope. This provides another way of plotting deviations in labeling ratio in terms of the residuals around the line of regression.