Is statistical parametric mapping valid for nonuniformly smooth biomechanical data?

It was recently reported that statistical parametric mapping (SPM) may be invalid for the analysis of nonuniformly smooth biomechanical data, like high frequency impact forces. The purpose of this study was to test that assertion. Using open-source software we simulated a variety of smooth Gaussian 1D data embodying a wide variety of nonuniform smoothness characteristics, and we computed Type I error rates across 10,000 simulation iterations for each type of nonuniform smoothness. Results showed that, in all cases, SPM accurately controlled Type I error rates at the prescribed alpha=0.05. Through separate simulations we speculate that the cited report may have either (i) inadvertently simulated non-Gaussian continua, or (ii) conflated region-of-interest error distribution with omnibus error, and that these errors, and not nonuniformity, yielded the reported departures from alpha=0.05. Our results suggest that SPM is, in fact, valid for nonuniformly smooth 1D data.