Parallel-line relative potency

What this model estimates

The model fits one intercept per preparation and a common slope on the selected log-dose axis. The horizontal separation between each test and reference line gives relative activity after back-transformation. Absolute potency is relative activity multiplied once by the assigned reference potency.

Use a dose region where the transformed response is adequately linear and the common-slope assumption is justified. This is a parallel-line potency calculation, not linear calibration of an unknown and not a raw-dose slope-ratio assay.

Data and design needed

Provide exactly one reference and at least one test preparation at a common grid of at least three distinct positive doses. Every preparation–dose cell must retain the same number of at least two replicates, numbered consecutively from 1, after any predeclared exclusions.

The design must identify the fitted lines and have finite, positive within-cell pure-error variance. Three levels are a software minimum, not evidence that an assay range, replication or linearity is scientifically adequate. Identical artificial replicates do not establish precision.

Check the settings

1. Check the reference identity, concentration unit, dose series and replicate map. On Model select Parallel-line (linear).
2. Choose Base-2 logarithm, Natural logarithm or Base-10 logarithm for Dose transformation. Untransformed dose is unsupported for this potency model.
3. Choose the justified Response transformation and Weighting. Response may be untransformed or use any of the three logarithms; all three weighting controls are supported within their domains. Inspect optional potency and interval-precision criteria before running.

Read the result and uncertainty

Inspect Model adequacy and comparability and Suitability and manual exclusions. In the PDF report, read each test’s relative activity/percentage, absolute potency and 95% interval together with the fitted settings. A larger test/reference ratio means greater potency on the declared dose basis.

Unweighted fitting uses ordinary least squares, pure-error F tests and a Student-t Fieller interval. Expected-response weighting uses the weighted fit, chi-square adequacy and a normal linearized Fieller interval. Both interval constructions retain numerator/denominator covariance; they are not an interval calculated from a slope alone.

Failures and limitations

Non-parallelism or non-linearity can fail suitability even when a numerical calculation completes. A significant preparation difference is not itself the fixed non-parallelism test. Read the target and meaning of each row.

Invalid logarithms or weights, an unbalanced design, zero pure error, a singular fit or an invalid Fieller interval cannot be repaired by accepting a substitute result. Check input errors, dose-region suitability and the experiment under your procedure. Do not remove observations merely to obtain parallelism.