impute.mdr: The Multifactor Dimensionality Reduction (MDR) Analysis for...

In imputeMDR: The Multifactor Dimensionality Reduction (MDR) Analysis for Incomplete Data

Description

This function provides various approaches to handling missing values for the MDR analysis of incomplete data to identify gene-gene interactions using biallelic marker data in genetic association studies

Usage

impute.mdr(dataset, colresp, cs, combi, cv.fold = 10, na.method = 0, max_iter = 30, randomize = FALSE)

Arguments

dataset	A matrix of SNP data with class variable (response; phenotype; disease status). Genotypes must be coded as allele counts (0,1,2). Missing genotypes should be coded as 3
colresp	Column number of class variables in the dataset. No missing value is allowed for the class variable
cs	The value used to indicate "case (affected)" for class variable
combi	The number of SNPs considered simultaneously as predictor variables (An order of interactions to analyze)
cv.fold	The number of folds k for k-fold cross-validation
na.method	Options for missing handling approaches. na.method = 0 for complete data, na.method = 1 for treating missing genotypes as another genotype category, na.method=2 for using available data for given number of SNPs under consideration as a model, na.method=3 for using method of imputing missing information by using EM (expectation-maximization) algorithm
max_iter	The number of maximum iteration in EM impute approach (na.method=3). In order to apply one-step EM approach, set this argument as 1
randomize	Logical. If 'TRUE' the cross validation sets are randomized

Value

min.comb	Marker combinations with the minimum error rate in each cross validation
train.erate	Training errors for selected marker combination
test.erate	Test error of the selected marker combination
best.combi	The best combination that was selected most frequently across k-fold cross validation

Author(s)

Junghyun Namkung, Taeyoung Hwang, MinSeok Kwon, Sunggon Yi and Wonil Chung

Maintainer: Junghyun Namkung <[e-mail]>

References

Namkung J, Elston RC, Yang JM, Park T. "Identification of gene-gene interactions in the presence of missing data using the multifactor dimensionality reduction method" Genet Epidemiol. 2009 Nov;33(7):646-56.

Examples

## sample data with missing values
data(incomplete)
## analysis example of 2nd order gene-gene interaction test
impute.mdr(incomplete, colresp=1, cs=1, combi=2, cv.fold = 10,na.method=2)

Example output

$cv.result
      SNP1   SNP2    train.err           test.err           
 [1,] "snp9" "snp10" "0.375438596491228" "0.392857142857143"
 [2,] "snp9" "snp10" "0.375"             "0.393939393939394"
 [3,] "snp9" "snp10" "0.372759856630824" "0.411764705882353"
 [4,] "snp9" "snp10" "0.382978723404255" "0.354838709677419"
 [5,] "snp9" "snp10" "0.387900355871886" "0.375"            
 [6,] "snp9" "snp10" "0.379432624113475" "0.354838709677419"
 [7,] "snp9" "snp10" "0.362007168458781" "0.5"              
 [8,] "snp9" "snp10" "0.371024734982332" "0.433333333333333"
 [9,] "snp9" "snp10" "0.368421052631579" "0.464285714285714"
[10,] "snp4" "snp6"  "0.393333333333333" "0.676470588235294"

$best
[1] "snp9"  "snp10"

imputeMDR documentation built on May 2, 2019, 12:03 p.m.