Each point is one full transient measurement at one heating power: the apparent diffusivity is shifted by self-heating, and the apparent value may increase or decrease with dissipated power depending on the material and temperature range — the platinum validation case of the companion article runs downward. The straight fit through the points, extended back to zero power (dashed), lands on the orange intercept; compare it against the green dashed line, the true unperturbed value. Flip the sensitivity slider between negative and positive to see both material families; raise the noise to see why a compressed power span extrapolates poorly.
Model: apparent diffusivity varying linearly with dissipated power, α(P) = α₀(1 + sP) with s of either sign, independent Gaussian noise per point and an ordinary least-squares fit extrapolated to P = 0. The slope is a power sensitivity under this configuration, not directly a temperature coefficient. The intercept standard error shown is the regression-derived value for this simplified model; a real campaign validates each point’s transient fit first. Axes are normalized; schematic teaching tool, not an instrument.