Global 4PL (shared parameters across datasets)

What is Global 4PL (shared parameters across datasets)?

When you have several dose-response curves on related drugs / cell lines / conditions, a global fit lets you share parameters across the curves rather than fit each independently. Common use: assume Top and Bottom are technical constants (shared) and let LogEC50 and HillSlope vary per drug.

The shared-parameter fit borrows strength: datasets that don't span the full dose range can pin down Top and Bottom by borrowing the asymptotes from datasets that do. Per-dataset fits would leave Top and Bottom unidentifiable for the partial-range curves.

When should I use Global 4PL (shared parameters across datasets)?

• Multi-drug dose-response panels.
• When you want to test whether Top or Bottom is genuinely constant across drugs.
• Compensating for partial dose-range coverage in some datasets.

What data does it need?

X + Y + dataset column + four shared-parameter checkboxes.

What does it report?

Single combined parameter table + per-dataset prediction curves.

What does it assume?

• Same 4PL model applies to all datasets.
• Shared parameters are genuinely constant — if they aren't, the fit will still converge but parameter estimates will be biased.

How do I interpret the result?

After the global fit, you can re-fit each dataset independently and compare via AIC to test whether the sharing assumption holds.

See also

• Sigmoidal dose-response (4PL)
• Compare two nonlinear models (F-test + AICc)