Glossary

Where the screen and the usual lab term differ, both are given.

A limit you set for one estimate, for example "Bias within ±2 mg/dL". Reports call it an acceptance criterion. See Set acceptance limits.

The substance or property measured, for example glucose. It provides a study filter and automatic headings in a project's overview. See Manage projects and colleagues.

The systematic difference between a result and a reference: an assigned value, comparative method, instrument or lot. Positive means the test reads higher. See Bias against an assigned value and Method comparison.

One saved calculation of one data version, with its estimates, outcomes and warnings. It never changes. See About records and history.

The fewest results needed to calculate a study, shown beside the suggested plan in the collection plan. See Set up.

The effect of one specimen on the next one measured, typically a high specimen raising the low one after it. See Carryover.

A result reported only as above or below a limit, for example <0.01. It is kept as entered. See Handle missing and censored values.

The panel at the top of Set up that shows how to run the study, with the suggested plan, the calculation minimum and the steps to follow. Select its heading to open it. See Set up.

The method the test method is compared with, plotted on the x-axis. It can have errors of its own. See Method comparison.

A range showing how precisely an estimate is known. See About confidence intervals and independent results.

The percentage attached to a confidence interval: 90%, 95% or 99%. A higher level gives a wider interval. See Uncertainty and denominators.

A change to a result. The original value stays on record, and the report's data listing shows both. See Correct or exclude a result.

The result of checking one estimate against one limit, shown as Met, Not met, Undecided, Not evaluated or Not applicable. The study as a whole then reads Criteria met, Criteria not met, Undecided or Not evaluated. See About acceptance decisions.

Coefficient of variation, CV% = 100 × SD / mean, the spread relative to the mean. It can't be calculated when the mean is zero or negative, or the scale has no true zero. See Uncertainty and denominators.

One saved version of a study's rows, settings and criteria, with who saved it. Each save that changes something makes a new version. See About records and history.

A regression allowing for error in both methods, using their error-variance ratio λ. Ordinary Deming assumes a constant SD; weighted Deming assumes a constant CV. See Method comparison.

A plot of each specimen's difference (test minus comparative), with the mean bias and 95% limits of agreement. Also called a Bland–Altman plot. See Method comparison.

Leaving a row out of the calculations, with a reason. The row stays in the study, exports and data listing. See Correct or exclude a result.

What is independently replicated and counted in n, such as a specimen, subject, material, pool or day. Repeats of one unit count once. See Uncertainty and denominators.

Results are independent when they share no source of error, for example different patients' specimens.

A substance that may change a measured result, for example hemoglobin from hemolysis. See Interference.

Cohen's kappa, the agreement between two qualitative methods beyond chance. It is descriptive and reported beside percent agreement. See Qualitative agreement.

A group of results analyzed separately, for example a control level. Some studies call it a partition or condition. See Set up.

Limit of blank, the highest result expected from a sample with no analyte, at a stated probability. See Detection limits.

Limit of detection, the lowest concentration reliably distinguished from the limit of blank. See Detection limits.

Limit of quantitation, the lowest concentration at which results meet a stated precision or accuracy goal. See Detection limits.

A control, calibrator, pool or other prepared sample.

A concentration where clinical decisions are made, for example 126 mg/dL for glucose. Method comparison estimates bias there. Outside the measured range the bias is extrapolated and can't meet a criterion. See Criterion outcomes.

The identifier of the calculation method behind a saved result. It changes when a change to the method can alter a value or a decision you see. Earlier results keep theirs. See Study reference.

How numbers are written, 1,234.56 or 1.234,56. You set it for each study. See Numbers, units and identifiers.

One entered result, one row. Several can come from one specimen or material. See Enter results in the grid.

A regression using the median of pairwise slopes. It resists outliers and makes no distribution assumption, but needs a positive linear relationship. See Method comparison.

The share of a population a reference interval should contain, for example 0.95. It is different from the confidence level. See Uncertainty and denominators.

Positive and negative percent agreement with a comparative method on its positive and negative results. See Qualitative agreement.

An optional, flat collection of related studies, with an overview, combined report and report-name settings. Studies started directly from the workspace go into General studies. See Manage projects and colleagues.

The range of results expected in a defined healthy population, usually its central 95%. It can be verified or established. See Reference interval verification and Reference interval establishment.

A method accepted as giving each specimen's true condition. Agreement with a reference standard gives sensitivity and specificity.

A fitted line relating the test method (y) to the comparative method (x). The options are ordinary least squares, Deming, weighted Deming and Passing-Bablok. See Method comparison.

Precision of one material measured repeatedly in one run, by one operator on one instrument. See Repeatability.

A repeated measurement of the same specimen or material. Repeats measure variation but do not add independent specimens. See how each study counts results.

The lowest to highest concentration a laboratory reports for a method. A linearity study verifies it with levels near each limit and criteria at every level between them.

Measurements made together under the same conditions, typically between calibrations or QC events.

Standard deviation, the typical spread around the mean, in the result's unit. Calculated as the sample SD (n − 1).

The share of truly positive and truly negative specimens a qualitative test identifies correctly. They need a reference standard. Without one, you get PPA and NPA. See Qualitative agreement.

The first stage of a study: its details, calculation settings and acceptance limits. Each method chapter describes them under Set up. See Set up: study details and settings.

sigma = (TEa% − |bias%|) / CV% at one concentration. An optional Minimum acceptable sigma gives a direct acceptance check; without it, the result is descriptive. See Sigma metric.

An optional action that records who reviewed a report: it fills in the name and date on page 1 and locks that version with its PDF. Without it, a report prints with blank Signed off by and Date lines to sign by hand. See Signing off.

A report locked by sign-off. Its HTML and PDF stay as signed. See About records and history.

A patient or prepared sample, identified by a coded specimen ID. It is the experimental unit in comparison studies.

The pass rule for method, instrument and reagent lot comparisons. Each specimen's difference is compared with its allowable difference. At least 95% of specimens must be within their limits. A method or instrument comparison needs at least 5 specimens. A reagent lot comparison needs at least 2 and a complete lot plan. A study that requires every specimen applies that rule at any count. See Method comparison and Lot comparison.

How long a specimen, reagent, calibrator or control stays within an allowable change from baseline under stated storage. See Stability.

One of the four parts every study moves through, shown as tabs: Set up, Data, Results and Report. Each tab shows its state: needs input, ready, done or out of date. See Finding your way.

One evaluation of one analyte with one design. It holds its data, settings, limits, calculations and reports. See Start a study.

A person contributing specimens to a reference interval study. Each subject counts once, even with several draws.

The practical starting design in a study's collection plan, taken from its Run this study guide: how many materials, levels, days, runs and replicates to collect, with the total. Adjust it to your question using the study guides.

Total allowable error, the largest error in one result a laboratory accepts at a concentration. Used by the sigma metric. See Sigma metric.

The method under evaluation, plotted on the y-axis. Differences are test minus comparative.

Precision across runs and days in one laboratory. See Multi-day precision.