mbmdr (version 2.6)

mbmdr.PermTest: Permutation Test for exploring significance of MB-MDR result.

Description

Performs a permutation test for specified interaction models from mbmdr object.

Usage

mbmdr.PermTest(x, n, model = NULL, sig.level=1)

Arguments

x

An mbmdr object returned by mbmdr function.

n

Number of permutations.

model

Vector specifiying an interaction model or matrix with rows referring to interaction models to be submitted for permutation testing. If model=NULL (default), permutation testing is performed on the best model derived from the x object.

sig.level

Significance level for the confidence intervals of the permutation based p-values, using a normal approximation (Nettleton, 2000). If sig.level=1 (as default), confidence intervals are not computed. When the permutation p-value is too small with respect to n, the normal aproximation is not appropiated and the C.I. are not provided (Nettleton, 2000).

Value

An object is returned of a new class, mbmdr.PermTest, with following attributes:

n

The number of permutations.

mbmdr

The mbmdr object used.

PermTest

A data.frame with results of the permutation tests. The following information is returned:

SNP1...SNPx
Names of snps for each specified interaction models.	NH
	Number of significant High risk genotypes in the interaction.
betaH
Regresion coeficient in step2 for High risk exposition.	WH
	Wald statistic for High risk category.
NL
Number of significant Low risk genotypes in the interaction.	betaL
	Regresion coeficient in step2 for Low risk exposition.
WL
Wald statistic for Low risk category.	Wmax
	Maximun Wald statistic for the interaction model.
Perm.P
Permutation p-value for the interaction model.	IC.lower
	Lower limit of the confidence interval for permutation p-value.
IC.upper
Upper limit of the confidence interval for permutation p-value.	SNP1...SNPx

Details

A permutation testing is performed for each specified model by permuting the outcome variable and calling the mbmdr function. The call to the mbmdr function is made by recovering the call from mbmdr object and replacing the outcome (by a permuted outcome vector) and SNP data (by the subset of the specified model). All other arguments in the initial call to mbmdr are transfered to mbmdr.PermTest.

References

Nettleton D., Doerge R.W. (2000) Accounting for Variability in the Use of Permutation Testing to Detect Quantitative Trait Loci. Biometrics, Vol. 56, No. 1, pp. 52-58.

See Also

mbmdr

Examples

  data(simSNP)
  fit <- mbmdr(y=simSNP$Y,data=simSNP[,3:12],order=2,list.models=c(2,1),
             family=binomial(link=logit))
             
  # Single model permutation test
  mbmdr.PermTest(fit,100)
  
  # Next steps takes some time
  
  # Permutation test for all models with MIN.P <= 0.05
  # order <- 2
  # models <- subset(fit$result, MIN.P <= 0.05, select = 1:order)
  # mbmdr.PermTest(fit,100,models)
  
  # Permutation test and confidence interval for all models with MIN.P <= 0.05 
  # mbmdr.PermTest(fit,100,models,sig.level=0.05)