twocor function

Confidence intervals for two-sided tests on correlation coefficients.

The twopcor function tests whether the difference between two Pearson correlations is 0. The twocor function performs the same test on a robust correlation coefficient (percentage bend correlation or Winsorized correlation).

twopcor(x1, y1, x2, y2, nboot = 599, ...)
twocor(x1, y1, x2, y2, corfun = "pbcor", nboot = 599, tr = 0.2, beta = 0.2, ...)

Arguments

• x1: a numeric vector.
• y1: a numeric vector.
• x2: a numeric vector.
• y2: a numeric vector.
• nboot: number of bootstrap samples.
• corfun: Either "pbcor" for percentage based correlation or "wincor" for Winsorized correlation.
• tr: amount of Winsorization.
• beta: bending constant.
• ...: currently ignored.

Details

It is tested whether the first correlation coefficient (based on x1 and y1) equals to the second correlation coefficient (based on x2 and y2). Both approaches return percentile bootstrap CIs.

Returns

twopcor and twocor return an object of class "twocor" containing:

• r1: robust correlation coefficient

• r2: value of the test statistic

• ci: confidence interval

• p.value: p-value

• call: function call

References

Wilcox, R. (2012). Introduction to Robust Estimation and Hypothesis Testing (3rd ed.). Elsevier.

See Also

pbcor, wincor

Examples

ct1 <- subset(hangover, subset = (group == "control" & time == 1))$symptoms
ct2 <- subset(hangover, subset = (group == "control" & time == 2))$symptoms
at1 <- subset(hangover, subset = (group == "alcoholic" & time == 1))$symptoms
at2 <- subset(hangover, subset = (group == "alcoholic" & time == 2))$symptoms

set.seed(111)
twopcor(ct1, ct2, at1, at2)
set.seed(123)
twocor(ct1, ct2, at1, at2, corfun = "pbcor", beta = 0.15)
set.seed(224)
twocor(ct1, ct2, at1, at2, corfun = "wincor", tr = 0.15, nboot = 50)