Figure 7.

Part 3–fMRI simulation on effects of model shift and duration. A: A library of reference hemodynamic waveforms was generated by altering two model parameters, the model shift, d, and the Model Duration, Δ, and convolving the task waveform with a hemodynamic response function. Model Shifts varied between 0–3 s while Model Durations were 1, 2, 5, 10, and 20 s. B: Correlation coefficient, r, curves are plotted as a function Model Shift, δ (error bars denote standard deviation). The maximum r is dependent on the optimal onset time, δ_optimal, and the curvature of the r curves illustrates the variability in sensitivity for the different Model Durations. Mismatch conditions, whereby the Model Duration is overestimated by 10% of the true duration are shown in gray. C: Evaluating r under the optimal condition (black) and under the mismatch condition (gray), i.e., 10% Δ mismatch and δ = δ_optimal + 1 s, shows a reduction in the expected correlation coefficient for brief tasks, but no effect within error for longer, block‐design tasks.