Sampling/ Power Analysis Overview
Power Analysis, Overview

Power Analysis: A Practical Overview

Power Analysis evaluates the probability a statistical test will detect a real effect. This guide explains the underlying statistics first, then shows how to apply them using AnalyZ Solutions.

8 min read Power Analysis Beginner friendly

Sample Size Estimation answers how many people do I need to estimate this value precisely. Power Analysis answers a related but different question: if I run this specific statistical test, how likely am I to detect a real effect, or how many people do I need to have a good chance of detecting it. This guide first explains what power means and how to choose it, then walks through how to apply it in AnalyZ Solutions.

01What statistical power and significance level mean

Power, formally 1 minus beta, is the probability that a study correctly detects an effect that is really there. A power of 80% means that if the effect genuinely exists at the size specified, there is an 80% chance the test comes back statistically significant, and a 20% chance of missing it entirely, purely due to sample size.

Significance level, alpha, is the counterpart risk. It is the probability of concluding an effect exists when it actually does not, a false positive. A significance level of 0.05, corresponding to 95% confidence, is the standard default across most fields.

How to choose these values

80% power and a 0.05 significance level are conventional defaults, not fixed rules. Consider raising power to 90% when missing a real effect would be especially costly, for example when the result determines whether a programme continues. Consider tightening significance level below 0.05 when a false positive would be especially costly, for example before recommending a policy change to a wide population based on the result. Relaxing either value below its convention is sometimes acceptable for lower stakes, exploratory work, but should be a deliberate choice rather than a default.

A no significant effect result is not the same as no effect. If a study was underpowered, a null result is nearly uninformative. It is equally consistent with there is no effect and there is a real effect, but the study could not detect it. A significance level set too loosely carries the opposite risk, an increased chance of concluding an effect is real when it is actually noise. This is why both power and significance level justification are expected before data collection, not calculated afterward to explain a disappointing or a surprising result. Once data collection is complete, there is no way to fix an underpowered study after the fact.

02Two ways to use each tool

Every test in Power Analysis works in either direction.

After calculating, a power curve chart shows how power changes across a range of sample sizes for small, medium, and large effects, a fuller picture than any single number can give.

03Choosing the right test

TestUse whenEffect size
Two-sample t-testComparing means between two independent groupsCohen's d
One-sample t-testTesting whether a group mean differs from a known reference valueCohen's d
Paired t-testComparing two measurements on the same individuals, pre and postCohen's dz
Chi-square testTesting association between two categorical variablesCohen's w
McNemar's testDetecting a change in a binary outcome measured twice on the same subjectsDiscordant pair proportions
One-way ANOVAComparing means across three or more independent groupsCohen's f
Linear regressionDetecting a given R² with one or more predictorsCohen's f²
Logistic regressionBinary outcome model with multiple predictorsOdds ratio and EPV
Paired versus independent, once more. If the same people are measured twice, use the paired t-test for a continuous outcome or McNemar's test for a binary outcome, not the two-sample or chi-square versions. Paired designs are almost always more efficient when the option is available.

04Understanding effect size

Every test requires specifying how large an effect should be detectable, expressed in standardised units so it is comparable across different outcome scales. Cohen's benchmarks are the field standard.

Effect sizeSmallMediumLarge
Cohen's d, t-tests0.20.50.8
Cohen's w, chi-square0.10.30.5
Cohen's f, ANOVA0.100.250.40
Cohen's f², regression0.020.150.35

When there is no prior basis for the effect size, no pilot data, no comparable published study, a medium effect is the standard planning default.

05Logistic regression's extra rule, EPV

Logistic regression has a second constraint beyond power, Events Per Variable, the number of outcome events divided by the number of predictors in the model. A conventional minimum of 10 events per predictor helps avoid an unstable, overfit model. The logistic regression tool reports both the power-based and EPV-based minimum sample size, and uses whichever is larger.

How to Use This in AnalyZ Solutions
  1. Choose the test matching your planned analysis. Select from the eight tests listed above.
  2. Choose achieved power or required sample size. Depending on whether a sample is already fixed or still being planned.
  3. Enter your significance level and effect size. Use a custom value from pilot data if available, or a Cohen benchmark otherwise.
  4. Calculate, and view the power curve. The chart shows how power changes across a range of sample sizes for small, medium, and large effects.
VIDEO WALKTHROUGH PLACEHOLDER, 90 SECONDS

A short screen recording showing these steps in the AnalyZ Solutions interface can be embedded here.

Frequently Asked Questions
How is this different from Sample Size Estimation?
Sample Size Estimation is generally built around estimating a value to a stated precision, or, for the Measuring Change designs, a specific real-world comparison such as an RCT or a before and after study. Power Analysis is built around named statistical tests run directly in analysis software, useful whenever a plan does not map cleanly onto one of the study-design tools.
What power should I target, 80% or 90%?
80% is the conventional minimum across most fields. Consider 90% for high stakes decisions where missing a real effect would be especially costly, such as a go or no-go funding decision.
Can I use a custom effect size instead of Cohen's benchmarks?
Yes. Every tool accepts a custom effect size value, not only the small, medium, and large presets. A pilot-derived estimate is always preferable to a generic benchmark when one is available.

Ready to check your study's power?

Open Power Analysis in AnalyZ Solutions. Free, browser based.

Try it out
Related Guides