Read suitability and uncertainty

Read the whole result

1. Check whether the calculation completed. If it failed, read the named design, domain, convergence or inference reason before interpreting the results.
2. Check suitability and each required criterion, including its target, observed value, limit and outcome. A number alone does not override a failed criterion.
3. For each preparation or sample, inspect the estimate, unit, interval availability and, for concentration, reportability and contributing observations. Read the exact evidence in the PDF report as well as the summary.

Relative-potency adequacy and optional limits

The fixed checks require evidence of regression (p < 0.05) and no detected non-parallelism or global non-linearity at that level (p ≥ 0.05). Preparation differences and per-preparation non-linearity are informational in the fixed partition. Failure to detect non-parallelism is not proof of biological equivalence.

On Model, optional criteria set a relative-potency range and confidence limits relative to the estimate for each test preparation. Interval precision compares lower limit / estimate and upper limit / estimate with their declared bounds; it is not a raw replicate-CV test. Specify limits and decision decimal places from your procedure. No universal assay acceptance range is supplied.

Unweighted parallel-line adequacy uses F tests and Student-t Fieller limits. Weighted parallel-line and the common-shape sigmoid paths use their fixed chi-square adequacy and normal linearized Fieller construction. The method is determined by the model/settings, not selected after seeing the result.

Concentration criteria and reportability

System suitability criteria offers Minimum calibration R² and optional recovery limits for individual reportable standard wells, individual QC wells and aggregated QC samples. A high R² alone does not establish model validity, sample recovery or adequate uncertainty. Standard recovery checks calibration observations; an independently prepared QC answers a different question.

Only concentrations inside the inclusive range derived from included reportable standards are eligible. An observation below or above range is censored, without a numeric extrapolation. If some observations remain eligible, the aggregate uses those observations: inspect the omitted ones too. With none eligible, a mixture of below- and above-range observations is indeterminate, not their numerical average.

Not configured means no such criteria were configured; it is not a suitability pass. A calculation may be reportable while failing a configured criterion. Follow your procedure for dilution, repeat measurement and investigation instead of treating a missing concentration as zero.

Understand the 95% interval

The interval expresses uncertainty under the selected calculation and its assumptions; it is not a specification range, a prediction of every future assay or a probability that a fixed true value lies in this particular interval. Missing preparation or biological variation is not created by the software. Dilution factors are treated as configured constants, without extra dilution uncertainty.

A sample with one usable observation keeps its observation’s interval; it does not use a multi-observation formula. An invalid or unbounded potency interval fails explicitly instead of being clipped or replaced. Concentration intervals can be unavailable with a reason; an unavailable interval is not zero uncertainty. Read the reported explanation and stop downstream use when your procedure requires an interval that is unavailable.

Keep display rounding separate from decisions

Fitting and saved results use unrounded calculations. Configured criteria use their saved decision precision and inclusive limits; ordinary screen formatting is separate. Read the decision value and exact bound before concluding that a rounded display contradicts a pass/fail result. Minimum calibration R² is configured to three decision decimal places in the current editor.

Predeclared manual exclusions require reasons and remain in the report. They occur before domain checks and fitting, and can make a design invalid or change the calibration range. There is no automatic outlier deletion, automatic replicate-CV rejection or silent refit with a different model.