corci: Compute a 1-alpha percentile bootstrap confidence interval...

In GRousselet/bootcorci: Bootstrap correlation confidence intervals

Description

The default correlation is the percentage bend correlation. When using Pearson's correlation, the confidence interval is adjusted to compensate for the error term's heteroscedasticity. Missing values are automatically removed.

Usage

corci(
  x,
  y,
  method = "pbcor",
  nboot = 2000,
  alpha = 0.05,
  alternative = "two.sided",
  null.value = 0,
  saveboot = TRUE,
  ...
)

Arguments

x, y	Two vectors of the same length.
method	A function that returns a correlation coefficient. Options in bootcorci include "pearson", "spearman", "pbcor", "wincor". Default is "pbcor".
nboot	Number of bootstrap samples. Default 2000.
alpha	Alpha level. Default 0.05. For corfun = pearson, alpha is restricted to 0.05 because the confidence interval adjustments have not been calculated for other alphas.
alternative	Type of test, either "two.sided", "greater" for positive correlations, or "less" for negative correlations.
null.value	Hypothesis to test. Default 0.
saveboot	Option to save bootstrap samples. Default TRUE.
...	Optional parameter to pass to correlation function.

Value

• estimate the correlation coefficient.

• p.value the p-value of the test.

• statistic the t statistic of the test.

• bootsamples the bootstrap samples.

Note

Modified from functions corb and pcorb from Rallfun-v37.txt - see https://github.com/nicebread/WRS/ and http://dornsife.usc.edu/labs/rwilcox/software/.

References

Wilcox, R.R. (2009) Comparing Pearson Correlations: Dealing with Heteroscedasticity and Nonnormality. Communications in Statistics - Simulation and Computation, 38, 2220–2234.

Wilcox, R.R. (2017) Introduction to Robust Estimation and Hypothesis Testing, 4th edition. Academic Press.

GRousselet/bootcorci documentation built on March 6, 2021, 7:13 a.m.