someMTP (version 1.4.1)

fdrOrd: Controlling the False Discovery Rate in ordered Test

Description

Ordinal procedure controlling the FDR and the Generalized FWER

Usage

fdrOrd(p, q = .01, ord = NULL, GD=FALSE)
kfweOrd(p, k = 1, alpha = 0.01, ord = NULL, alpha.prime = alpha, J = qnbinom(alpha, k, alpha.prime), GD = FALSE)

Arguments

p

vector of p-values

ord

Values on the basis of which the procedure select the hypotheses (following decreasing order). The vector have the same length of p. If NULL the natural ordering is considered.

q

average FDR level

alpha

global significance level

k

number of allowed errors in kFWE controls

J

number of allowed jumps befor stopping

alpha.prime

univariate alpha for single step Guo and Romano procedure

GD

Logic value. Should the correction for general dependence be applied?

Value

• The function returns an object of class someMTP.object.
• rej:
  a logical vector indicating whenever the related hypotesis have been rejected.
• p:
  the vector of p-values used in the call
• ord:
  The vector used to sort the p-values (decrasing).
• MTP:
  "fdrOrd" or "kfweOrd"
• GD:
  A logical value incating if the correction for General Dependence have been used or not.
• q:
  The level of controlled FDR.
• alpha:
  The level of controlled k-FWER
• alphaprime:
  The significance level of individual tests
• k:
  Number of allowed Errors
• J:
  Number of allowed Jumps

References

L. Finos, A. Farcomeni (2010). k-FWER Control without p-value Adjustment, with Application to Detection of Genetic Determinants of Multiple Sclerosis in Italian Twins. Biometrics.

See Also

See also draw

Examples

set.seed(17)
	x=matrix(rnorm(60),3,20)
	x[,1:10]=x[,1:10]+2 ##variables 1:10 have tests under H1
	ts=apply(x,2,function(x) t.test(x)$statistic)
	ps=apply(x,2,function(x) t.test(x)$p.value) #compute p-values
	m2=apply(x^2,2,mean)           #compute ordering criterion

	pOrd <- fdrOrd(ps,q=.05,ord=m2)   #ordinal Procedure
	pOrd
	draw(pOrd)	
	sum(p.adjust(ps,method="BH")<=.05)  #rejections with BH
	
	kOrd <- kfweOrd(ps,k=5,ord=m2)                          #ordinal procedure
	kOrd
	kOrdGD <- kfweOrd(ps,k=5,ord=m2,GD=TRUE)                  #ordinal procedure (any dependence)
	kOrdGD