Control statistics
13 min
- Calculate the mean of control measurements
- Calculate standard deviation from stated control limits and state its unit
- Calculate coefficient of variation from a mean and standard deviation
- State the share of control results expected within 1, 2, and 3 SD of the mean
Try first
Get the idea
Four numbers
A control material is measured over and over, and its results spread around a central value. Four numbers describe those results, and every QC rule is built on them.
- The mean is the arithmetic average: add the results and divide by how many there are.
- The standard deviation (SD) says how far results usually sit from the mean. It carries the analyte's own unit, such as mmol/L or mg/dL.
- The variance is the SD squared.1
- The coefficient of variation (CV) removes the unit, so levels and methods can be compared:
CV (%) = SD ÷ mean × 100
Because the SD keeps its unit, a bigger SD at a higher concentration can still be the better performance.
SD from the limits
A control sheet often lists only the limits. Limits of ±2 SD run from 2 SD below the mean to 2 SD above it.
- The mean is the midpoint.
- The whole interval is 4 SD wide, so 1 SD is the width divided by 4.
- Dividing by 2 gives an SD twice the true one and draws every limit too far out.2
What each band means
When the results of a stable procedure follow a Gaussian distribution, fixed shares fall inside each band:3
| Band around the mean | Results inside | Results outside |
|---|---|---|
| ±1 SD | 68.3% | 31.7% |
| ±2 SD | 95.4% | 4.6% |
| ±3 SD | 99.7% | 0.3% |
Those shares explain how the rules treat each band:4
- A control beyond 2 SD turns up about once in every 22 results from a procedure that is working. A run with two control levels gives two chances, so about 9% of good runs show one. That is why a single result beyond 2 SD serves as a warning to check the other rules.
- A result beyond 3 SD happens by chance about 3 times in 1,000, rare enough to reject the run.
Laboratories set each control's mean and SD from their own repeated measurements of one control lot, collected over many runs. The examples here use a handful of results to show the arithmetic. A handful is far too few to show whether a spread is Gaussian.2
References
- National Institute of Standards and Technology. Measures of scale. NIST/SEMATECH e-Handbook of Statistical Methods. Accessed September 23, 2026.
- Clinical and Laboratory Standards Institute. Statistical Quality Control for Quantitative Measurement Procedures: Principles and Definitions. 4th ed. CLSI guideline C24Ed4E. Clinical and Laboratory Standards Institute; 2016.
- National Institute of Standards and Technology. What do we mean by "normal" data?. NIST/SEMATECH e-Handbook of Statistical Methods. Accessed September 23, 2026.
- Westgard JO. Westgard rules and multirules. Westgard QC. Accessed September 23, 2026.
Watch one
A new sodium control lot arrives with laboratory-established limits of 147.0 to 153.0 mmol/L, stated as ±2 SD. Find the mean, the SD, the CV and the ±3 SD limits.
- Take the midpoint: (147.0 + 153.0) ÷ 2 = 150.0 mmol/L.
The ±2 SD limits sit the same distance on either side of the mean.
- Divide the width by 4: (153.0 − 147.0) ÷ 4 = 6.0 ÷ 4 = 1.5 mmol/L.
From the lower limit to the upper limit is 2 SD down plus 2 SD up, 4 SD in all.
- Divide the SD by the mean: 1.5 ÷ 150.0 × 100 = 1.0%.
The CV turns the SD into a share of the mean, so its only unit is %.
- Add and subtract 3 SD: 150.0 − 4.5 = 145.5 mmol/L and 150.0 + 4.5 = 154.5 mmol/L.
Every other limit on the chart is a whole number of SDs from the mean.
Your turn
Results
- Calculate the mean of control measurements
- Calculate standard deviation from stated control limits and state its unit
- Calculate coefficient of variation from a mean and standard deviation
- State the share of control results expected within 1, 2, and 3 SD of the mean
To review
3 questions from this step will come back in Review.
Next step: Reading a Levey-Jennings chartReview nowOpen the part
Keep
Sources checked
The rest of this step
A short briefing, a demonstration at the bench and 3 practice problems.
A free account opens the rest and keeps your progress.