Skip to content
Omnikalc

Compare two correlations calculator (Fisher z)

Enter two correlation coefficients and their sample sizes to get the Fisher z test of the difference, its p-value and the 95 % confidence interval of each correlation.

Updated on
Your inputs

Value out of range

Value out of range

Value out of range

Value out of range

Results

z statistic of the difference

1.825

Two-sided p-value
0.0680
No significant difference between the two correlations
Fisher z of correlation 1
0.7250
Fisher z of correlation 2
0.3654
Lower 95 % limit of correlation 1
0.413
Upper 95 % limit of correlation 1
0.766
Lower 95 % limit of correlation 2
0.105
Upper 95 % limit of correlation 2
0.555
Difference in Fisher z with its 95 % margin
0.3862

The compare two correlations calculator tests whether two correlation coefficients from independent samples differ, with the Fisher z test, its p-value and the 95 % interval of each correlation.

How the calculation works

  • Enter the two correlations and their sample sizes.
  • Read the z statistic and the p-value.
  • Check the confidence intervals and the margin on the difference.

Example

A correlation of 0.62 is found in 50 people and 0.35 in 60 other people. The Fisher values are 0.7250 and 0.3654. The standard error of the difference is √(1/47 + 1/57) = 0.197, so z = 1.825 and the two-sided p-value is 0.0680. The 95 % interval of the first correlation is 0.413 to 0.766, and of the second 0.105 to 0.555.

Why it matters

The calculator helps psychology, education and health researchers who compare relationships between groups, students of statistics, and analysts who check whether an effect changed between two periods.

Good to know

The test assumes independent samples and roughly bivariate normal data, and it works best with at least 25 people per sample. A non-significant difference does not prove the correlations are equal, especially with small samples, since the test has low power. Very different correlations can still have overlapping intervals, so compare with the test and not by eye. For correlations measured on the same people, a different test is needed. To compare more than two correlations, use a chi-square test on the Fisher values weighted by n − 3. Remember that a correlation is not a causal effect. Report the sample sizes, the confidence intervals and the way each variable was measured, so that readers can judge whether the two groups are really comparable.

The correlation significance calculator tests one correlation against zero, the partial correlation calculator removes a third variable, the Spearman correlation calculator uses ranks and the confidence interval calculator brackets means.

Frequently asked questions

Why use the Fisher z transformation?

The sampling distribution of a correlation is skewed, especially near ±1. Fisher’s transformation z = ½ ln((1 + r) / (1 − r)) is approximately normal with a standard error of 1 / √(n − 3), whatever the true correlation, which allows tests and intervals.

How are two correlations compared?

Transform both, subtract them and divide by √(1 / (n1 − 3) + 1 / (n2 − 3)). The result is compared with the standard normal distribution. The samples must be independent: it is not valid for two correlations from the same people.

How are the intervals built?

The limits are computed on the z scale as z ± 1.96 / √(n − 3), then converted back with r = (e^(2z) − 1) / (e^(2z) + 1). That is why the intervals are not symmetric around r.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.