Statistical Power Analysis for Pearson Correlation
Also known as: Korelasyon Güç Analizi, power analysis for r, sample size for correlation
Correlation power analysis is a pre-study calculation that determines how many participants are needed — or how much statistical power an existing sample provides — for a Pearson correlation test. Formalised by Jacob Cohen in his landmark 1988 text, it uses the expected correlation coefficient r directly as the effect size, so researchers can plan studies that are neither underpowered nor wastefully large.
Key highlights
- Uses the expected correlation coefficient directly as the effect size, avoiding indirect transformations required by other power methods.
- Grounded in Cohen's (1988) widely accepted conventions, making results immediately interpretable across disciplines.
- Applicable to both one-tailed and two-tailed tests, accommodating directional and non-directional hypotheses.
- Computationally lightweight; yields a precise closed-form answer rather than requiring simulation.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use correlation power analysis whenever you plan a study whose primary goal is to estimate or test a Pearson correlation between two continuous variables. Three conditions should be met: both variables are continuous and approximately normally distributed, the relationship is expected to be linear, and you can provide a plausible prior estimate of r (from published literature, a pilot study, or Cohen's conventions). The analysis is appropriate for cross-sectional survey studies, experimental pre/post designs, and psychometric validation studies. With fewer than ten participants the calculation becomes unreliable; a pilot study to refine the effect size estimate is strongly recommended before finalising the sample plan.
Strengths & limitations
- Uses the expected correlation coefficient directly as the effect size, avoiding indirect transformations required by other power methods.
- Grounded in Cohen's (1988) widely accepted conventions, making results immediately interpretable across disciplines.
- Applicable to both one-tailed and two-tailed tests, accommodating directional and non-directional hypotheses.
- Computationally lightweight; yields a precise closed-form answer rather than requiring simulation.
- Sensitive to the accuracy of the prior r estimate; an optimistic or inflated pilot correlation leads to systematic underestimation of the required sample size.
- Assumes bivariate normality and a linear relationship; if the true relationship is non-linear the planned power may not materialise.
- Does not account for planned covariates, partial correlations, or attrition; additional adjustments are needed for complex designs.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
Where do I get the expected r if I have no prior data?
The safest starting point is a systematic review or meta-analysis of similar studies in your field. If no prior evidence exists, use Cohen's (1988) medium-effect convention of r = 0.30 as a conservative planning value. Avoid basing the estimate on a single small pilot study, as pilot correlations are highly variable.
Should I use a one-tailed or two-tailed test?
Use a two-tailed test by default; it does not require you to commit to a direction in advance and is the standard expectation in most journals. A one-tailed test is only appropriate when the direction of the correlation is specified a priori based on strong theoretical grounds, and it reduces the required sample size by concentrating the rejection region on one side.
Why is 80 % power the conventional target?
Cohen (1988) proposed 80 % as a pragmatic balance between the costs of data collection and the risk of a false negative. At alpha = 0.05 and power = 0.80 the ratio of a Type II error to a Type I error is 4:1, which Cohen judged acceptable for most behavioural-science applications. Higher-stakes decisions may warrant 90 % or 95 % power.
What if my study involves a partial or semi-partial correlation?
Standard correlation power analysis applies only to a zero-order Pearson r. For partial or semi-partial correlations, or for correlations tested within a regression framework, you should use regression-specific power analysis tools that account for the additional predictors.
Sources
- 1.Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Lawrence Erlbaum Associates.ISBN 978-0805802832
You have read it. What now?
Cite this page
ScholarGate. (2026, June 1). Correlation Power Analysis. ScholarGate. https://scholargate.app/statistics/power-analysis-correlation