Choose a scientific model

Start with the quantity you need

Relative potency compares a test preparation with a reference preparation across dose levels. Concentration determination uses known standards to estimate an unknown concentration from its response. A common-shape 4PL potency analysis and a 4PL calibration answer different questions, even though both use logistic curves.

Use your procedure and prior assay characterization to justify the family, curve shape, response scale and variance assumption. A rising response, an assay name, a high R-squared or an assistant’s suggestion alone does not establish the appropriate model. Do not repeatedly switch models until one passes.

Check that the design can identify the model

These counts are minimum requirements for calculation, not a sufficient assay design or acceptance recommendation. Replication, dose placement, range and signal variation matter. Relative-potency models additionally require one reference, at least one test, a common positive dose grid and the same count of at least two replicates in every preparation–dose cell after exclusions.

Concentration models need identifiable standards and a reportable interval supported by included reportable levels. An anchor can aid fitting outside that interval; it neither establishes curve shape nor becomes reportable by being included. For 5PL calibration, at least two distinct levels must lie on each side of the fitted maximum-slope transition.

Select and inspect the model

1. In development, confirm the analysis family, import and map the plate, then inspect the concentrations and replicate assignments.
2. On Model, select the intended card. Review the model-specific page below, then inspect transformations, Weighting and, for concentration, Curve orientation. A logarithmic concentration axis does not by itself justify response weighting.
3. Review the suitability criteria from your procedure, wait for Draft saved and choose Run analysis. Check the model name, version and settings alongside the result. The scientific model version is separate from the product release.

When to stop

Clarify the scientific question and experimental design before choosing settings. If the data do not support the model, check for input errors and review the design under your laboratory procedure. You may need a different dose range or a repeat experiment.