Increasing your sample size increases statistical power, which is the probability that your study will detect a real effect when one exists. Power is calculated as 1 minus the probability of a Type II error (a false negative), so a study with 80% power has a 20% chance of missing a true effect. Sample size is one of four interconnected factors that determine power, alongside effect size, significance level, and the type of statistical test used. Understanding how these pieces fit together helps you design studies that can actually answer the question you’re asking.
What Statistical Power Actually Measures
Statistical power answers a simple question: if there’s a real difference or relationship in what you’re studying, how likely is your test to find it? A study with low power might fail to detect a genuine treatment effect, leading you to wrongly conclude that nothing is going on. This is a Type II error, sometimes called a false negative.
The standard threshold in most fields is 80% power, meaning you accept a 20% chance of missing a real effect. Some clinical trials aim for 90% power when the stakes are higher. The 2025 CONSORT guidelines for reporting randomized trials consistently reference these benchmarks. For example, one trial calculating sample size at 80% power needed 202 participants, while another aiming for 90% power with a 20% loss-to-follow-up allowance required 214 participants (107 per group).
Why Larger Samples Increase Power
The core mechanism is straightforward. When you measure something in a small group, random variation has an outsized influence on your results. A few unusual values can shift your average substantially, making it hard to tell whether a pattern in your data reflects something real or just noise. As your sample grows, those random fluctuations cancel each other out. Your estimate of the true average (or true difference between groups) becomes more precise, and the spread of uncertainty around that estimate shrinks.
This shrinking uncertainty is what drives the power increase. With less noise obscuring the signal, your statistical test becomes more sensitive to genuine effects. Think of it like listening for a specific voice in a crowd: the more recordings you have to compare, the better your ability to pick out that voice from background chatter.
The Four Factors That Determine Power
Sample size doesn’t work in isolation. Four variables are tightly linked, and changing any one of them shifts the others:
- Sample size: More participants means more precision and higher power.
- Effect size: The magnitude of the difference or relationship you’re trying to detect. Larger effects are easier to find and require fewer participants.
- Significance level (alpha): The threshold for calling a result “statistically significant,” typically set at 5%. Lowering alpha (making it harder to declare significance) reduces power unless you compensate with a larger sample.
- Type of statistical test: Different tests have different sensitivity to patterns in data, which affects the sample size needed.
These four factors create a balancing act. Reducing the probability of one type of error increases the risk of the other. Setting a very strict significance level protects against false positives but makes false negatives more likely, unless you add more participants to compensate.
Effect Size Changes Everything
The size of the effect you’re looking for dramatically changes how many participants you need. A large, obvious effect (say, a drug that cuts symptom severity in half) can be detected with a relatively small study. A subtle effect (a drug that improves outcomes by 5%) requires a much larger sample to distinguish from random variation.
This is why small sample sizes combined with small effect sizes are the most common recipe for underpowered studies. If you’re investigating a treatment with a modest expected benefit, you need proportionally more participants to have a reasonable chance of detecting it. Researchers planning a study typically estimate the expected effect size from prior research, then calculate the sample size needed to achieve 80% or 90% power at that effect size. Software tools like G*Power are widely used for this calculation, taking effect size, desired power, significance level, and the planned statistical test as inputs.
Diminishing Returns at Large Sample Sizes
The relationship between sample size and power is not a straight line. Early increases in sample size produce large jumps in power. Going from 20 to 60 participants might boost power from 40% to 80%. But the gains taper off as you approach the upper end. Going from 200 to 400 participants might only move power from 95% to 99%.
This curve means there’s a practical sweet spot. Below it, your study is too weak to be informative. Above it, you’re spending resources (time, money, participant burden) for marginal improvement. The goal of a power analysis before starting a study is to find this sweet spot: the minimum sample size that gives you adequate power for the effect size you expect.
What Happens When Studies Are Underpowered
Running a study with too few participants doesn’t just risk a null result. It raises serious scientific and ethical concerns. An underpowered study lacks the scientific validity to produce reliable estimates of treatment effects, which means the time, funding, and effort invested are largely wasted. Worse, participants in clinical trials are exposed to the risks and burdens of the study with little chance that their contribution will yield meaningful knowledge.
The Journal of the Royal Society of Medicine has characterized otherwise well-designed but underpowered clinical trials as unethical, arguing that they divert participants and resources from properly designed studies that could actually answer the research question. Ethics review boards in the US are expected to assess scientific validity, including whether a study’s sample size is adequate, before approving it. If investigators know or suspect at the outset that their study will be too small to produce valid data, failing to disclose this to participants in the informed consent process is considered a form of deception.
Beyond ethics, underpowered studies contribute to misleading bodies of evidence. When a small study fails to find an effect, it’s tempting to conclude the treatment doesn’t work. But the study may simply have been too small to detect a real benefit. This is why reporting power calculations in published research (including effect size assumptions and the target power level) is a requirement in major reporting guidelines like CONSORT.
How to Use This in Practice
If you’re designing a study, run a power analysis before collecting any data. You need three inputs to calculate the required sample size: your desired power level (80% is standard, 90% for high-stakes work), your significance level (usually 5%), and an estimate of the expected effect size. The effect size estimate typically comes from pilot data or published studies on similar questions.
If you’re reading a published study, check whether the authors report a power analysis. A study that found “no significant difference” between groups is only informative if it had enough power to detect a meaningful difference in the first place. Without that information, a negative result tells you very little. Look for the reported sample size, the target power, and the assumed effect size to judge whether the study was adequately designed to answer its own question.