| Hypothesis test for equality of two correlation coefficients {corrfuns} | R Documentation |
Hypothesis test for equality of two correlation coefficients
Description
Hypothesis test for equality of two correlation coefficients.
Usage
correls2.test(r1, r2, n1, n2, type = "pearson")
Arguments
r1 |
The value of the first correlation coefficient. |
r2 |
The value of the second correlation coefficient. |
n1 |
The sample size of the first sample from which the first correlation coefficient was computed. |
n2 |
The sample size of the second sample from which the first correlation coefficient was computed. |
type |
The type of correlation coefficients, "pearson" or "spearman". |
Details
The test statistic for the hypothesis of equality of two correlation coefficients is the following:
Z=\frac{\hat{z}_1-\hat{z}_2}{\sqrt{1/\left(n1-3\right)+1/\left(n2-3\right)}},
where \hat{z}_1 and \hat{z}_2 denote the Fisher's transformation (see correl applied to the two correlation coefficients and n_1 and n_2 denote the sample sizes of the two correlation coefficients. The denominator is the sum of the variances of the two coefficients and as you can see we used a different variance estimator than the one we used before. This function performs hypothesis testing for the equality of two correlation coefficients. The result is the calculated p-value from the standard normal distribution.
Value
The test statistic and its associated p-value for the test of equal correlations.
Author(s)
Michail Tsagris
R implementation and documentation: Michail Tsagris mtsagris@uoc.gr.
See Also
Examples
y <- rnorm(40)
x <- matrix(rnorm(40 * 1000), ncol = 1000)
a <- correls(y, x )