Coefficient of Variation

The coefficient of variation (CV) is a standardized measure of dispersion calculated as the ratio of the standard deviation to the mean, typically expressed as a percentage. It allows comparison of the relative variability of measurements across different scales, units, or populations where direct comparison of standard deviations would be misleading.

Researchers frequently need to compare the variability of measurements that are on different scales or have different magnitudes. The standard deviation alone is not suitable for such comparisons because it is expressed in the original units of measurement. The CV solves this problem by expressing variability relative to the mean, producing a unit free measure that enables meaningful cross variable and cross population comparisons. In laboratory science, the CV is used to assess the precision and reproducibility of analytical methods. In clinical research, it quantifies the variability of biological measurements within and between individuals. In manufacturing and quality control, it measures process consistency.

The CV is calculated by dividing the standard deviation by the mean and multiplying by 100 to express the result as a percentage. A CV of 10% indicates that the standard deviation is 10% of the mean.

The CV is most meaningful when the variable is measured on a ratio scale (one with a true zero point), such as concentration, weight, or duration. For variables measured on an interval scale (one without a true zero, such as temperature in Celsius or IQ scores), the CV is not meaningful because the ratio to the mean depends on the arbitrary zero point.

In laboratory science, the intra assay CV measures the variability of repeated measurements within a single analytical run, while the inter assay CV measures the variability across multiple runs. Both are used to assess the precision of analytical methods, with lower CVs indicating greater consistency. Acceptable CV thresholds vary by application; in clinical chemistry, CVs below 5% are typically considered good precision, while in immunoassays, CVs below 15% may be acceptable.

The CV can also be used to compare the variability of different measurements in the same population. For example, a researcher might compare the CV of systolic blood pressure with the CV of heart rate to determine which measurement is more variable relative to its mean.

The CV is appropriate for comparing variability across measurements with different units or different scales, for assessing the precision and reproducibility of analytical methods, for evaluating the consistency of manufacturing processes, and for comparing biological variability across different physiological measurements. It is not appropriate for variables with means close to zero (where the CV becomes unstable and grows without bound as the mean approaches zero), for variables measured on interval scales without a true zero, or for variables with negative values.

A common error is calculating the CV for variables where the mean is near zero or negative, which produces misleading or undefined values. Another mistake is comparing CVs across samples with very different sample sizes without considering that the CV is more precise with larger samples.

Researchers sometimes also confuse the CV with the standard deviation and use it in contexts where the standard deviation is the appropriate measure of variability (such as when all measurements are on the same scale and have similar means). In laboratory method validation, failing to distinguish between intra assay and inter assay CVs can lead to incomplete assessment of precision. Finally, reporting the CV without the underlying mean and standard deviation prevents readers from fully interpreting the measure.

A clinical biochemist validates a new enzyme immunoassay for cortisol by running 20 replicates of three quality control samples (low, medium, and high concentration). The intra assay CVs are 8.2%, 5.4%, and 4.1% for the low, medium, and high samples, respectively. The inter assay CVs, calculated from 15 independent runs over three weeks, are 12.3%, 7.8%, and 5.9%. The higher CVs at low concentrations reflect the greater relative variability when measuring near the assay's lower limit of detection. Against the laboratory's pre specified acceptance limit of 10% for both intra assay and inter assay CVs, the biochemist concludes that the assay meets the precision criteria at medium and high concentrations but not at the low concentration (inter assay CV 12.3%), where results may require repeat testing.