The relationship between mean channel selection and the calculated coefficient of variation

Abstract

Calculated coefficients of variation (CV) taken from the quotient of the standard deviation (S.D.) and the mean value of measured distributions are often used as an indicator of system performance in linear flow cytometry (FCM). The ability of the calculated CV to estimate the true CV of the underlying experiment before grouping (channelization) is dependent on the relationship between the width of the data channels and the magnitude of the S.D. of the measured distribution. When the channel width is equal to the S.D. of a distribution, the calculated CV is approximately 20% larger than the true CV of an experiment. By the time the S.D. is only one-half of a channel width, the calculated CV is unreliable. When the distribution S.D. is narrower than a channel's width, small changes in the distribution mean value will cause large variations in the calculated CV. As the true CV decreases, the calculation must be made with higher mean channel values. This dependence of calculated CV accuracy upon the relationship between S.D. and channel width places limitations upon mean channel selection that must be considered when using CV calculations for evaluating system performance, especially when looking for small improvements during optical alignment procedures. When an instrument is assumed to have a constant CV and the data are collected linearly, it is possible to improve the CV estimation accuracy by placing distributions in higher-numbered channels.(ABSTRACT TRUNCATED AT 250 WORDS)