Statistical Terminology

Statistical Terms

This glossary defines statistical terms common to laboratory quality control and necessary to understanding how to use proficiency testing data.

Coefficient of Variation (CV). Standard deviation expressed as a percentage of the mean; derived by dividing the standard deviation by the mean and multiplying by 100.

               CV = SD ÷ Mean X 100

Confidence Limits. Percentiles of the expected distribution of assays. Assuming a Gaussian distribution, 95% of sequentially performed tests on a control specimen will fall within two standard deviations of the mean.

Upper and lower control limits-established by applying two standard deviations to the mean-represent a confidence limit of95%.

Control Sample. May be called "control specimen" or "control." Specimen with a known concentration for specific analytes used to ensure that test systems are functioning properly. Acceptable ranges for control results may be established by the manufacturer or by individual testing facilities.

Duplicates. A simple check oflaboratory precision by dividing a specimen and processing the two portions independently, preferably not in sequence. After completion of a given determination, the two values are compared to see how closely they approximate one another.

Gaussian Distribution Curve. Bell-shaped curve produced as a result of plotting control values from consecutive runs of similar tests on a single control specimen. The values are plotted on the x-axis; the frequency of each value on the y-axis. The narrower the curve, the more precise the test.

Lot. A specific quantity of a control sample and/or reagent that has unifom1 character and quality within specified limits, produced in batches according to a single order during the same manufacturing cycle. All determinations that make up your quality control statistics should be of the same lot. When it is necessary to change the lot for your controls, establish a new mean value and standard deviation and begin a new quality control gTaph. Do this while still using the prior control.

Mean, Arithmetic. May be called "average" or "mean." Value obtained by dividing the sum of several results by the number of results.

                               Sum of Results ÷ Number of Results = Mean

Median Value. The middle value in an ordered set, or list of values, below and above which there are an equal number of values.

Precision. Degree to which repeated analyses of the same material approximate one another.

Shift. Quality control values that fall on the same side of the mean for six or more consecutive days with no tendency toward a consistent rise or fall; usually caused by modifications to procedures, either intentional or unintentional.

Standard Deviation. The difference between an individual value and the arithmetic mean is a deviation. Broadly speaking, the standard deviation represents the average of the individual deviations.

Standard Deviation Index (SDI). Denotes the difference between a test result and the group mean. The calculation of the SDI normalizes the bias of your result and, therefore, allows for a comparison of results from specimens having different concentrations of an analyte.

SDI = Your Mean - Group Mean ÷ Group SD

Trend. Quality control values that continue to increase or decrease over six or more consecutive days; usually caused by deterioration and/or evaporation of reagents, gradual increase in pH, or gradual reagent or instrument contamination.