Contact: Salimetrics (USA)
Power Analysis Tool
Salimetrics · Study Planning
Salivary Study Power Analysis
A planning and education tool to help investigators think through study design assumptions before working with a biostatistician. Not intended for use in grant proposals.
For planning and educational purposes only
Use this tool to: explore design assumptions, understand the relationship between assay precision and sample size, and prepare informed questions for your biostatistics consultant. Then take those assumptions to an independent statistician for the final, citable calculation.
This tool is not intended to serve as the power analysis in a grant proposal — it is designed to help investigators think through study design assumptions and understand how parameters like effect size, sample size, and assay precision interact.
This tool is not intended to serve as the power analysis in a grant proposal — it is designed to help investigators think through study design assumptions and understand how parameters like effect size, sample size, and assay precision interact.
Background: statistical power and study design
Power is the probability that your study will detect a real effect if one truly exists. A study with 80% power has an 80% chance of finding a statistically significant result and a 20% chance of missing a real effect entirely — producing a null result that won't be publishable and won't answer your research question.
Effect size (Cohen's d)
How large is the difference you expect to detect? Small (0.2), medium (0.5), or large (0.8). When in doubt, use medium — it's the most common in salivary biomarker research.
Statistical power (1 – β)
The probability of detecting the effect if it exists. 80% is the standard academic threshold. For high-stakes or confirmatory studies, 90% or 95% is recommended.
Alpha (α)
The false positive rate you're willing to accept. 0.05 is standard — it means a 5% chance of declaring an effect that isn't real. Use 0.01 for stricter inference.
Assay precision (CV%) and its effect on power
Coefficient of variation (CV%) measures how consistently an assay produces the same result on repeated measurements of the same sample. High CV inflates measurement noise, which competes with the biological signal you are trying to detect. A CV of 20% can require 30–50% more participants to achieve the same power as a CV of 5%.
Your study parameters
Enter your planned or current total enrollment. For between-subjects designs, this is split evenly across groups.
If you have a prior study or pilot, use that effect size. Otherwise, 0.5 is a reasonable default for salivary biomarker studies.
Most funding agencies expect at least 80% power. For NIH R01s, 80–90% is standard.
Use 0.05 unless your field or funder requires stricter inference.
Within-subjects designs are more efficient — fewer participants needed for the same power.
Add a buffer for participants who may not complete the study. 10–20% is typical for longitudinal designs.
Leave as default if using Salimetrics. Substitute the QC metrics for assays performed by another laboratory.
Select your study parameters above
Fill in all fields and your power analysis results will appear here automatically.
Quick reference glossary
Statistical power (1 – β)
The probability of detecting a real effect. 80% power means a 20% chance of missing a true difference.
Effect size (Cohen's d)
A standardized measure of the magnitude of a difference. Small = 0.2, medium = 0.5, large = 0.8.
Alpha (α)
The probability of a false positive — declaring an effect that doesn't exist. Standard is 0.05.
Type II error (β)
Missing a real effect. Equal to 1 – power. At 80% power, β = 0.20.
Between-subjects
Different participants in each condition. Requires more participants than within-subjects designs.
Within-subjects
Same participants measured across conditions or timepoints. More statistically efficient.
Attrition buffer
Extra participants enrolled to account for expected dropout. Common in longitudinal and clinical designs.
Minimum detectable effect
The smallest effect size your study can reliably detect given your sample size, alpha, and power target.