Common-shape 4PL relative potency

What this model estimates

The preparations share the lower and upper asymptotes and slope while their horizontal locations differ. Those horizontal shifts estimate potency relative to the reference. The symmetric logistic shape is defined on the fitted response and logistic dose coordinate, not necessarily the original plotted axes.

This common-shape potency model fits increasing responses. It is not a set of independent four-parameter fits whose midpoint ratios are substituted for a joint potency result. Separate shape diagnostics help assess the common-shape premise; they do not replace it.

Data and design needed

Use exactly one reference and at least one test, at least five common positive dose levels, and the same count of at least two consecutively numbered replicates per preparation–dose cell after exclusions.

Plan enough dose coverage to identify both response plateaus and the changing part of the curve, with informative overlap between preparations. Meeting the five-level minimum alone does not identify asymptotes, slope or potency. Within-cell pure-error variance must be finite and positive.

Check the settings

1. Inspect the reference, units, mapped doses and replicates, then select Common-shape 4PL on Model.
2. Dose transformation supports Untransformed positive doses or Base-2, Natural or Base-10 logarithm. Response transformation supports Untransformed or the same logarithms. Inspect the Weighting assumption and domain.
3. Review procedure-defined potency and interval-precision criteria and the saved state. Do not select 4PL solely because its graph looks smoother.

Read the result and uncertainty

Read each test’s relative activity, percentage/absolute potency and 95% Fieller limits in the PDF report. Inspect the model diagnostics and fixed regression, non-parallelism and global non-linearity checks on the result page.

All weighting choices use the model’s final-linearized normal Fieller interval, with replicated-cell pure-error variance and joint horizontal-shift covariance. Nonlinear curve-parameter covariance is separate from the covariance used for potency inference. An R-squared value alone is not a replacement for these checks.

Failures and limitations

Insufficient tail coverage, weakly determined plateaus, invalid weight domains, failure to converge, zero pure error or invalid confidence limits can prevent a usable result. A completed fit can still fail suitability or its common-shape assessment.

Use the scientific reason and input evidence to decide the next action. The model does not fall back to parallel-line or 5PL, and a failed adequacy check is not an instruction to delete a well.