The uncertainty values you provide describe the precision of your measurements. Choose whether their magnitudes should be used as given or treated as nominal values whose common scale is estimated from residuals (the differences between data and model). This choice changes parameter uncertainties, not the fitted parameters. The solver, fitting weights, residuals and goodness of fit are unchanged. Model uncertainty bands and Model Evaluator uncertainties can change.

Absolute — Do Not Scale Parameter Uncertainties: Treat the provided values as independent 1σ measurement uncertainties and propagate them to the parameters without adjusting their scale to match the residual scatter. Absolute does not mean exact; constant and percentage expressions can both specify absolute uncertainties.

Nominal — Scale Parameter Uncertainties Using Residuals: Use the provided values to set the relative precision of the measurements, then estimate their common scale from the residuals. Parameter uncertainties are multiplied by , so they may increase or decrease. This is curve.fit’s historical convention, including custom equations, and is the default for all models, including Weighted Mean.

For the usual propagated uncertainty of a weighted average, select Absolute. With one measurement, Nominal cannot estimate the uncertainty from residuals; Absolute can propagate the supplied uncertainty. Without active measurement uncertainties, only Nominal is available: uncertainty must be estimated from residual scatter. Follow the uncertainty convention specified by your course or experiment. See Help for details, examples and assumptions.