Journal: International Biomechanics

Article Title: Statistical parametric mapping of biomechanical one-dimensional data with Bayesian inference

doi: 10.1080/23335432.2019.1597643

Figure Lengend Snippet: Classical SPM{ t } results for the three datasets. Top row shows descriptive statistics for each dataset (Mean ± 1 SD error cloud). Bottom row shows the frequentist inferences. The horizontal dashed lines depict the critical t * based on α = 0.05 and RFT calculations of residual smoothness. Supra-threshold clusters result in p < 0.05. For the GaitSymmetry example, note that these are time series from a single subject, the SD-cloud thus represents within-subject variability instead of between-subject variability. The inference only pertains to this subject

Article Snippet: Paired-sample t -test GaitSymmetry [198 x 101] , Classical SPM (Python): 0.025 s Bayesian SPM (RStudio): 1.580 s.

Techniques:

Journal: International Biomechanics

Article Title: Statistical parametric mapping of biomechanical one-dimensional data with Bayesian inference

doi: 10.1080/23335432.2019.1597643

Figure Lengend Snippet: Classical SPM{t} results for the three datasets. Begin and end-points of supra-threshold cluster locations are given as a percentage of the total movement time

Article Snippet: Paired-sample t -test GaitSymmetry [198 x 101] , Classical SPM (Python): 0.025 s Bayesian SPM (RStudio): 1.580 s.

Techniques:

Journal: International Biomechanics

Article Title: Statistical parametric mapping of biomechanical one-dimensional data with Bayesian inference

doi: 10.1080/23335432.2019.1597643

Figure Lengend Snippet: Overview of supra-threshold clusters for the Bayesian SPM tests (interval H 0 only). The less conservative q * = 0.05 threshold always yields broader clusters than the P ( H | data)* = 0.95 threshold. For the SimulatedTwoLocalMax and PlantarArchAngle datasets, the difference between both thresholds is small. For the GaitSymmetry example, the difference is larger and results in 4–7 separate clusters or 2 broad clusters (for the ultra-wide setting, it is nearly 1 cluster over the entire time span). The GaitSymmetry example also shows sensitivity to the scale of the prior, whereas this sensitivity was negligible in the other two datasets

Article Snippet: Paired-sample t -test GaitSymmetry [198 x 101] , Classical SPM (Python): 0.025 s Bayesian SPM (RStudio): 1.580 s.

Techniques:

Journal: International Biomechanics

Article Title: Statistical parametric mapping of biomechanical one-dimensional data with Bayesian inference

doi: 10.1080/23335432.2019.1597643

Figure Lengend Snippet: (a) Classical power (omnibus) analysis for calculating the number of trials necessary to reject H 0 given alpha = 0.05 and a minimal 2° difference between the left and right leg ( H 1 ). Horizontal lines show the typical power criterions of 0.80 and 0.90. Panels (b) and (c) give the classical and Bayesian SPMs using the first n = 55 gait cycles (for which classical power >0.80). For the Bayesian SPM, the maximal posterior error probability for which q < 0.05 was 0.113

Article Snippet: Paired-sample t -test GaitSymmetry [198 x 101] , Classical SPM (Python): 0.025 s Bayesian SPM (RStudio): 1.580 s.

Techniques:

Journal: International Biomechanics

Article Title: Statistical parametric mapping of biomechanical one-dimensional data with Bayesian inference

doi: 10.1080/23335432.2019.1597643

Figure Lengend Snippet: Computation time for frequentist and Bayesian SPM tests

Article Snippet: Paired-sample t -test GaitSymmetry [198 x 101] , Classical SPM (Python): 0.025 s Bayesian SPM (RStudio): 1.580 s.

Techniques: T-Test