Related Experiment Videos
The effect of trial size and variability on statistical power
J S Dufek1, B T Bates, H P Davis
1Department of Exercise and Movement Science, University of Oregon, Eugene 97403-1240.
Medicine and Science in Sports and Exercise
|February 1, 1995
Summary
Computer modeling reveals that larger sample sizes and trial sizes significantly boost statistical power. Single-subject (SS) experiments generally show poorer results than group analyses, particularly with smaller mean differences.
Area of Science:
- Statistics
- Biostatistics
- Computer Modeling
Background:
- Statistical power is crucial for detecting true effects in research.
- Understanding factors influencing statistical power is essential for robust experimental design.
- Single-subject (SS) research designs have unique considerations compared to group designs.
Purpose of the Study:
- To investigate the impact of sample size, mean differences, and subject variability on statistical power using a validated computer model.
- To compare statistical power between single-subject (SS) and group study results.
- To analyze the influence of different model complexities on statistical outcomes.
Main Methods:
- Development and validation of a computer model incorporating Monte Carlo procedures.
- Simulation of varying sample sizes (subjects and trials), mean differences, and subject variability.
- Comparison of results from simple and complex models (MOD1, MOD4) for both group and SS analyses.
Main Results:
- Increased mean differences positively correlated with improved statistical power.
- Group analyses yielded higher power (63.6%-100%) than SS analyses (16.8%-100%), especially with smaller mean differences.
- Higher subject variability in complex models (MOD4) reduced significant results and the Complex/Simple ratio for group analyses, but favored complex models in SS analyses.
Conclusions:
- Sample size and trial size are major determinants of statistical power.
- Group analyses generally offer superior statistical power compared to SS analyses.
- Subject variability and model complexity interact significantly, impacting results differently across SS and group designs, highlighting the importance of careful consideration in SS experiments.