Permutation Tests

DESCRIPTION:

Perform permutation tests for observations from specified data and for specified statistics.

USAGE:

permutationTest(data, statistic, B = 999, ..., 
                resampleColumns, ... 
                alternative = "two.sided", 
                combine = F, 
                combinationFunction = combinePValues.Fisher) 

REQUIRED ARGUMENTS:

data
data to be bootstrapped. May be a vector, matrix, or data frame.
statistic
statistic to be bootstrapped; a function or expression that returns a vector or matrix. It may be a function which accepts data as the first argument; other arguments may be passed using args.stat.
Or it may be an expression such as mean(x,trim=.2). If data is given by name (e.g. data=x) then use that name in the expression, otherwise (e.g. data=air[,4]) use the name data in the expression. If data is a data frame, the expression may involve variables in the data frame. For examples see .

OPTIONAL ARGUMENTS:

B
number of random permutations to use. With the default value of B=999, p-values are multiples of 1/1000.
resampleColumns
numerical, logical, or character, for subscripting columns of the data. If supplied, then only those columns of the data are resampled. This is useful for permutation tests; for example, for a permutation test of the correlation between two variables, permute only one of them.
...
additional arguments, which are passed to . For more detail on these arguments see
alternative
one of "two.sided", "greater", or "less". May be a vector of the same length as the result of calling statistic.
combine
numerical, logical, or character vector, indicating which components of the result to use for computing combined p-values. Or this may be a list, each of whose elements indicate a set of variables to use.
combinationFunction
a function which combined p-values; see help for combinePValues.Fisher for specifications

VALUE:

an object of class "permutationTest" which inherits from "resamp"

DETAILS:

This is implemented as a front end to , using as the sampler.

REFERENCES:

Pesarin, F. (2001), Multivariate Permutation Tests with Applications to Biostatistics, Wiley, Chichester, UK. (Describes nonparametric combination methodology.)
We wish to thank Dr. Luigi Salmaso for help designing and testing this function.

SEE ALSO:

, , . The latter two are for permutation tests comparing two groups.

More details on arguments, see . Note that calls , so many of the arguments are common.

Combination of p-values for multivariate statistics: , , , .

Print, summarize, plot: , , , ,

Description of a "permutationTest" object, extract parts: , , , .

Modify a "permutationTest" object: .

For an annotated list of functions in the package, including other high-level resampling functions, see: .

EXAMPLES:

# Test significance of a correlation 
y <- rnorm(30); x <- 1:30 
plot(x,y); abline(lm(y ~ x)) 
obj <- permutationTest(y, cor(y, x)) 
obj 
print(obj, digits=2) 
plot(obj)