A computational framework for estimating statistical power and planning hypothesis-driven experiments involving...

Area of Science:

• Biomechanics
• Statistical analysis
• Experimental design

Background:

• Statistical power is crucial for hypothesis-driven research.
• Limited methods existed for power assessment in continuum and one-dimensional (1D) time series data before the mid-1990s.

Purpose of the Study:

• To describe continuum-level power analyses for planning hypothesis-driven biomechanics experiments using 1D data.
• To demonstrate 1D effect modeling for sample-size calculations in single- and multi-subject experiments.

Main Methods:

• Theory-driven power analysis using the minimum jerk hypothesis for single-subject reaching experiments.
• Pilot-driven power analysis utilizing a published knee kinematics dataset for between-subject analysis.

Main Results:

• Achieved statistical power of approximately 0.8 with small sample sizes (five for within-subject, ten for between-subject).
• Sample size determination is contingent on a priori biomechanical meaning and effect size justifications.

Conclusions:

• The proposed technique encourages a priori justification of 1D effects, structuring experimental inquiry.
• Shifts research focus from seeking statistical significance to identifying non-rejectable hypotheses.