calcPvalsMC {perm}R Documentation

Internal functions: Calculate p-values from Complete Enumeration or Monte Carlo

Description

These internal functions are not meant to be called directly. This help is to explain the functions. The function calcPvals calculates p-values from a set of either completely enumerated test statistics (in which case all wgts are equal to one) or a set of test statistics from a network algorithm (in which case the wgts come from the network algorithm). The function calcPvalsMC calculates p-values and the associated confidence intervals from a Monte Carlo simulation. The p-values are estimated by (S+1)/(N+1) where S is the number of more extreme test statistics, and N is the number of Monte Carlo simulations. This estimator is guaranteed to be valid, and the confidence interval is the appropriate one associated with this estimator (see e.g., Fay, Kim, and Hachey, 2007).

Usage

calcPvalsMC(Tj, T0, digits, alternative = NULL, twosidedTstat = FALSE, p.conf.level=0.99)
calcPvals(Tj, T0, digits, denom = NULL, wgts = NULL, twosidedTstat = FALSE)
pCI(S, N, p.conf.level = 0.99)

Arguments

Tj numeric vector of values of test statistic
T0 observed value of test statistic
digits number of significant digits for rounding (see details)
alternative character vector: one of "less", "greater", "two.sided"
twosidedTstat logical, TRUE denotes test statistic is twosided
denom total number of test statistics
wgts vector of weights the same length as Tj, NULL gives all ones
S number of Monte Carlo replications equal or more extreme than T0
N number of Monte Carlo replications
p.conf.level confidence level for confidence intervals about Monte Carlo p-values

References

Fay, MP, Kim, H-J, Hachey, M (2007). Journal of Computational and Graphical Statistics. 16:946-957. Available at http://www3.niaid.nih.gov/about/organization/dcr/BRB/staff/michael.htm

See Also

Called from permTS, permKS, permTREND


[Package perm version 0.9-1.3 Index]