5PL calibration

What this model estimates

The model adds asymmetry to the lower asymptote, upper asymptote, location and slope of a logistic calibration. Eligible sample responses are inverted, returned to concentration units and dilution-corrected. This page describes scientific model version v2. Check the model version when comparing earlier results.

Choose it when justified by the assay relationship and a design that can estimate asymmetry. More freedom can expose weak identification; the application does not select 5PL automatically because a 4PL result failed.

Data and design needed

At least six distinct included standard levels are required. At least two distinct levels must lie below and two above the fitted maximum-derivative transition. Cover the informative changing region and tails; six arbitrary levels do not guarantee an identifiable 5PL curve.

Declare reportable and anchor standards, units, replicate treatment and sample dilution. The reportable interval still comes only from included reportable standards. An anchor’s position does not itself demonstrate asymmetric shape.

Check the settings

1. Select 5PL calibration and the scientifically intended increasing or decreasing Curve orientation.
2. Inspect Concentration transformation and Response transformation: Untransformed or Base-2, Natural or Base-10 logarithm. Choose one of the three justified weighting options within its valid domain.
3. Review the derived range, configured R²/recovery criteria and saved state. The model’s fitting limits and checks for a well-determined curve are fixed. They are separate from the assay acceptance criteria you set.

Read concentrations and the shape assessment

Read each unknown/QC concentration with reportability, interval and suitability, then inspect the fitted curve, standard recovery, asymmetry and diagnostic evidence in the saved result and report.

The v2 fit checks agreement of competing fitted solutions, finite full-rank information, transition coverage and a connected, finite 95% asymmetry profile strictly inside its computational bounds. The profile may include asymmetry 1; it tests identifiability, not automatic selection of 4PL.

Accepted observation estimates use 95% normal delta uncertainty with curve and response terms. Multiple eligible observations use the shared-curve, declared response-group log-concentration aggregate construction. The asymmetry profile is a separate shape diagnostic, not the sample concentration interval.

Failures and limitations

Weak identifiability, insufficient transition coverage, an unbounded asymmetry profile, weight nonconvergence or invalid inversion can prevent a usable result even when a plotted curve seems plausible. Read the displayed explanation to decide whether to correct an input or investigate the experimental design.

Provenarium keeps the selected 5PL model and does not extrapolate beyond the reportable range. Correct any confirmed input errors or review the design and repeat the work under your laboratory procedure.

• Scientific background: Gottschalk and Dunn (2005), 5PL characterization