Sample size and statistical power in [15O]H2O studies of human cognition | JoVE Visualize

Sample size and statistical power in [15O]H2O studies of human cognition

¹Mental Health Clinical Research Center, University of Iowa College of Medicine, Iowa City 52242, USA.

Journal of Cerebral Blood Flow and Metabolism : Official Journal of the International Society of Cerebral Blood Flow and Metabolism|September 1, 1996

View abstract on PubMed

Summary

Determining optimal sample size for Positron Emission Tomography (PET) studies is vital. A sample size of 10-20 subjects ensures good replicability in human cognition research, balancing detection with minimal false negatives.

Area of Science:

Neuroimaging
Cognitive Neuroscience
Biostatistics

Background:

Sample size determination is critical for Positron Emission Tomography (PET) studies.
Traditional power calculations face challenges with complex PET data analysis, including exploratory strategies and correlated network data.
Assessing sample size requires considering the unique demands of PET imaging analysis.

Purpose of the Study:

To investigate the impact of varying sample sizes on the replicability of results in human memory PET studies.
To establish a reliable sample size range for cognitive neuroscience research using PET.
To evaluate the primary criterion of replicability across different sample sizes.

Main Methods:

Examined effects of sample size (n=33, 16/17, 11, 6/7) in a human memory PET study.
Assessed replicability using hierarchical parameters: peak pattern, location, number, size, and t/z statistic intensity.
Used a large sample (n=33) as the gold standard for comparison.

Main Results:

Decreasing sample size led to an increase in false negatives and some loss of pattern/peak detection.
No corresponding increase in false positives was observed with smaller sample sizes.
Good replicability was achieved with sample sizes of 10-20 subjects.

Conclusions:

A sample size of 10-20 subjects is recommended for human cognition PET studies.
This range ensures good replicability for paired subtraction comparisons with 4-13% blood flow differences.
Findings provide guidance for optimizing sample size in neuroimaging research.