Run this study#
Your question: how far is our average result from this material's assigned value?
Bring: two materials at low and high concentrations, with their assigned values, units and any stated standard uncertainty. Suggested starting plan: 5 measurements a day for 5 days = 25 results per material, 50 for two materials. CLSI's public EP15 overview describes bias estimation from at least 25 measurements over at least five days.
- Set the acceptable bias first. Choose New study, pick Bias check and choose Create study. On Set up, under Acceptance limits, choose Enter your own limit for Bias, in the study unit, as a percentage of the assigned value, or both, or Use a CLIA / CAP limit when its analyte, fluid, method and unit fit your study (see CLIA / CAP limits). Check that each assigned value applies to your material, method and unit.
- Collect the readings. Follow the material's preparation and storage instructions. Measure under routine conditions with acceptable QC. You can reuse the results of a multi-day precision study on the same materials.
- Enter each material separately. Add a Level column for more than one material, and use the same material ID for every replicate. Record Day and Run for each reading: days
1to5, with run1on each day for this plan. On Data, choose Import a file and choose your file (start from Blank template (CSV) or accuracy.csv). Answer any question the dialog asks, then choose Import into this study. Enter one assigned value per level on Set up, under Assigned values, with its uncertainty when known. - Calculate and review. Choose Calculate results. Results opens. Check 25 results per level, the assigned value and the sign of the bias. Review each limit under Why, then on Report choose Download PDF.
What you get: bias, relative bias and recovery against the assigned value at each tested concentration; these decide the limits. With the 5-day plan the bias has no confidence interval, because repeats of one material on different days aren't independent results. Replicates measured in one run add intervals for that run (see Statistics). Calculation minimum: two results per level for a limit decision.
Purpose#
For each level the study compares the mean with the level's assigned value:
- Bias: mean minus assigned value, in the study unit.
- Relative bias: bias as a percentage of the assigned value.
- Recovery: mean as a percentage of the assigned value.
When to use it#
Use it for a certified reference material, a proficiency-testing sample with a peer-group mean, or a sample valued by a comparison laboratory. For patient specimens measured by two methods use Method comparison. For a dilution series use Linearity or Dilution verification.
Study setup#
Bias check sits under Comparison and bias in the New study dialog, which also asks for the Assigned value. That value is for level 1.
On Set up, the Assigned values table holds one row per level. Add the Standard uncertainty if you have it. For more levels, once the data are in, choose Add assigned-value rows for N levels in the data (shown while some levels have no assigned value), then enter each value. Each level is analysed separately.
| Field | Required? | What to enter |
|---|---|---|
| Level | Yes | The level name used in the data, for example 1 |
| Assigned value | Yes | A number in the study unit |
| Standard uncertainty | No | Standard uncertainty (k = 1) in the study unit |
Entering data#
On Data, choose Import a file or Add row.
| Column | Required? | What to enter |
|---|---|---|
| Material ID | Yes | For example CRM-1 |
| Result | Yes | Each measured value |
| Level | No | For more than one material. Must match a level in Assigned values. Default 1 |
| Day | Recommended | The collection day, for example 1 to 5. Left blank if the file has no Day column |
| Run | Recommended | The run within that day, for example 1. Left blank if the file has no Run column |
Correct or exclude a censored value such as <5. See Missing and censored values.
Every level with results needs an assigned value before you calculate. See Calculating results.
Statistics#
For each level:
| Estimate | How it is calculated |
|---|---|
| Bias (mean minus target) | Mean − assigned value, in the study unit |
| Relative bias | 100 × bias / |assigned value|, in % |
| Recovery | 100 × mean / assigned value, in % |
| Mean, Sample SD / CV | Arithmetic mean. Sample SD (n − 1). 100 × SD / mean |
| Target standard uncertainty | The stated u, in the study unit |
| Measurement bias CI | Mean CI − target, with mean CI = mean ± t(1 − α/2, n − 1) × SD / √n |
| Combined standard uncertainty | u_c = √(SD²/n + u_target²), for independent observations only |
| Bias CI including target uncertainty | Bias ± t(1 − α/2, n − 1) × u_c, when u_c and the measurement bias CI are available |
Relative bias and recovery need a positive assigned value on a ratio scale. Otherwise the result gives the reason they weren't calculated.
Which intervals a level gets depends on how its results relate. Mean, bias, relative bias and recovery are always calculated and decide the limits.
| Results at a level | Material ID | Day and Run | Mean CI, bias CI and combined uncertainty |
|---|---|---|---|
| Replicates of one material in one run | The same on every row | The same on every row | Calculated from the replicate SD, t(n − 1). They describe that run's repeatability only |
| Repeats of one material across days or runs, such as the 5-day plan | The same on every row | More than one day or run | Not given: results from one material on different days aren't independent, and SD / √n would understate the uncertainty |
| Repeats of one material without a recorded day or run | The same on every row | Blank on some rows | Not given, as above |
| Separate, independent specimens that share one assigned value | Different on every row | Any | Calculated from all results as independent results |
Give a different ID only to a separate specimen. Replicates of one material keep its ID, even when that means no interval.
Acceptance limits for this study#
- A Bias limit compares each level's mean minus its assigned value. A percentage is taken of that level's own assigned value.
- Limits on Relative bias and Recovery, in
%, are under More statistics. - A value that can't be calculated, or a level with 1 included result, gives Undecided.
- If you correct an assigned value, a Mean or Median limit built on the old value becomes Undecided. Edit the limit on Set up, save and recalculate.
One failing limit fails the whole study. See Acceptance limits and how outcomes are decided.
Worked example#
accuracy.csv holds ten results for one reference material, all with material ID CRM-1, Day 1 and Run 1, and no level column. Because they are one run, the intervals below describe that run's repeatability. The same material measured over 5 days, as in the suggested plan, gives the bias without intervals.
Setup: analyte Glucose, unit mg/dL, 95% confidence. Assigned value for level 1: 100.0 mg/dL, standard uncertainty 0.15 mg/dL. Limits: Bias (-1 to 1 mg/dL) and Recovery (98 to 102 %).
What the saved result shows#
| Estimate | Displayed value |
|---|---|
| Included results | 10 |
| Mean | 100.10 mg/dL |
| Sample SD / CV | 0.258 mg/dL / 0.26% |
| Assigned target | 100 mg/dL |
| Target standard uncertainty | 0.15 mg/dL |
| Bias (mean minus target) | +0.10 mg/dL |
| Relative bias | +0.1% |
| Recovery | 100.1% |
| 95% measurement bias CI | −0.08 to 0.28 mg/dL |
| 95% bias CI including target uncertainty | −0.29 to 0.49 mg/dL |
| Criterion | Observed | Acceptance limit | Outcome |
|---|---|---|---|
| Bias | +0.10 mg/dL | ±1 mg/dL | Met |
| Recovery | 100.1% | 98–102% | Met |
The study status is Criteria met.
The example project's study Glucose · assigned-target bias check uses the 20 replicates of repeatability.csv against 100 mg/dL: mean 100.17 mg/dL, bias +0.17 mg/dL, relative bias +0.2%, recovery 100.2%. Its Bias limit, ±2 mg/dL, is met.
Reading the results#
- Results and acceptance shows the limits, then one row per level with the assigned target, mean, bias, relative bias, recovery and both intervals.
- Additional statistics and calculation history → Assigned-target bias check adds the target's uncertainty.
- Bias keeps its sign: positive means your method reads high. 0.1 % relative bias is the same as 100.1 % recovery.
- Measurement bias CI covers replicate scatter. Bias CI including target uncertainty adds the target's uncertainty and is wider.
When a limit fails#
- Recheck the assigned value: a typo, the wrong level, or an expanded uncertainty entered as standard (divide it by its coverage factor).
- Recheck data entry: the imported columns, decimal convention, a stray replicate.
- If the bias is real, check for an expired or mishandled CRM, calibration drift or a reagent lot change, and rerun with a fresh aliquot.
Statistical detail#
Sources: NIST/SEMATECH e-Handbook §1.3.5.2 and §1.3.5.6. See Methods and sources.