🧪 This platform is in early beta. Features may change and you might encounter bugs. We appreciate your patience!
underpowered_study
An underpowered study has too few participants or observations to reliably detect an effect of the expected size. Statistical power is the probability that a study will detect a true effect when one exists. Studies with power below 80% (a common convention) are considered underpowered. Such studies produce unreliable results: significant findings are likely inflated in magnitude, and non-significant findings cannot be interpreted as evidence of no effect.
A study with 15 participants per group tests whether a new therapy reduces depression symptoms. The expected effect size requires 80 participants per group for 80% power. The study finds p = 0.08 and concludes 'no significant effect.' This does not mean the therapy does not work; the study simply lacked the sample to detect it.
A startup tests its mindfulness app on 12 employees to see if it reduces workplace stress. The study finds a small improvement that doesn't reach statistical significance (p = 0.11) and concludes the app 'shows no effect,' when in reality the sample was far too small to detect a plausible benefit.
Researchers investigate whether a rare genetic variant is associated with a neurological condition by recruiting 20 affected individuals and 20 controls. They find no significant association and publish a null result, but the study had only 15% power to detect the expected effect — the absence of evidence is mistaken for evidence of absence.
Binary (yes/no) questions an LLM must answer to identify this aspect:
Was a power analysis conducted before the study to determine needed sample size?
Type: binaryIs the sample size adequate for the expected effect size?
Type: binaryAre null findings being interpreted as 'no effect' without considering low power?
Type: binaryIs a significant finding from a small sample being treated as definitive?
Type: binaryAn underpowered study has too few participants or observations to reliably detect an effect of the expected size. Statistical power is the probability that a study will detect a true effect when one exists. Studies with power below 80% (a common convention) are considered underpowered. Such studies produce unreliable results: significant findings are likely inflated in magnitude, and non-significant findings cannot be interpreted as evidence of no effect.
Sample size calculations are technical and rarely reported in the media or in press releases. Audiences assume that any published study is adequately sized, treating non-significant results as definitive null findings.
Check the reported sample size against the expected effect size. If the study is small and finds no effect, note that it may be underpowered. Look for power analyses in the methods section.
Underpowered studies are common in neuroscience, pilot clinical trials, and social science experiments. Button et al. (2013) found the median statistical power of neuroscience studies was just 21%.
Equal measurement error across groups that typically biases estimates toward the null.
The tendency to draw strong conclusions from small samples, failing to recognize that small samples are more variable and less reliable than large ones.
Believing that small samples accurately represent the underlying population distribution.
Bayesian and frequentist approaches yield contradictory conclusions with large sample sizes.
Use these tools to detect, analyze, or train this aspect.