iterativeBMA (version 1.30.0)

crossVal: Cross Validation for Iterative Bayesian Model Averaging

Description

This function performs k runs of n-fold cross validation on a training dataset for survival analysis on microarray data, where k and n are specified by the user.

Usage

crossVal(exset, survTime, censor, diseaseType="cancer", nbest=10, maxNvar=25, p=100, cutPoint=50, verbose=FALSE, noFolds=10, noRuns=10)

Arguments

exset

Data matrix for the training set where columns are variables and rows are observations. In the case of gene expression data, the columns (variables) represent genes, while the rows (observations) represent samples. The data is not assumed to be pre-sorted by rank.

survTime

Vector of survival times for the patient samples in the training set. Survival times are assumed to be presented in uniform format (e.g., months or days), and the length of this vector should be equal to the number of rows in exset.

censor

Vector of censor data for the patient samples in the training set. In general, 0 = censored and 1 = uncensored. The length of this vector should equal the number of rows in exset and the number of elements in survTime.

diseaseType

String denoting the type of disease in the training dataset (used for writing to file). Default is 'cancer'.

nbest

A number specifying the number of models of each size returned to bic.surv in the BMA package. The default is 10.

maxNvar

A number indicating the maximum number of variables used in each iteration of bic.surv from the BMA package. The default is 25.

p

A number indicating the maximum number of top univariate genes used in the iterative bic.surv algorithm. This number is assumed to be less than the total number of genes in the training data. A larger p usually requires longer computational time as more iterations of the bic.surv algorithm are potentially applied. The default is 100.

cutPoint

Threshold percent for separating high- from low-risk groups. The default is 50.

verbose

A boolean variable indicating whether or not to print interim information to the console. The default is FALSE.

noFolds

A number specifying the desired number of folds in each cross validation run. The default is 10.

noRuns

A number specifying the desired number of cross validation runs. The default is 10.

Value

• The output of this function is a series of files giving information on cross validation results. The file beginning with 'foldresults' contains information for every fold in the form of a 2 x 2 table indicating the number of test samples in each category (high-risk or censored, high-risk or uncensored, low-risk or censored, low-risk or uncensored). This file also gives the accumulated percentage of uncensored statistic from each run. The file beginning with 'runresults' gives the total number of test samples assigned to each category along with percentage uncensored across the entire run. The end of this file contains this same information, averaged across all runs. The file beginning with 'stats' gives the statistics from each fold, including the p-value, chi-square statistic, and variance matrix. Finally, the file beginning with 'avg_p_value_chi_square' gives the overall means and standard deviations of the p-values and chi-square statistics across all runs and all folds.

Details

This function performs k runs of n-fold cross validation, where k and n are specified by the user through the noRuns and noFolds arguments respectively. For each run of cross validation, the training set, survival times, and censor data are re-ordered according to a random permutation. For each fold of cross validation, $1/n$th of the data is set aside to act as the validation set. In each fold, the iterateBMAsurv.train.predict.assess function is called in order to carry out a complete run of survival analysis. This means the univariate ranking measure for this cross validation function is Cox Proportional Hazards Regression; see iterateBMAsurv.train.wrapper to experiment with alternate univariate ranking methods. With each run of cross validation, the survival analysis statistics are saved and written to file.

References

Annest, A., Yeung, K.Y., Bumgarner, R.E., and Raftery, A.E. (2008). Iterative Bayesian Model Averaging for Survival Analysis. Manuscript in Progress.

Raftery, A.E. (1995). Bayesian model selection in social research (with Discussion). Sociological Methodology 1995 (Peter V. Marsden, ed.), pp. 111-196, Cambridge, Mass.: Blackwells.

Volinsky, C., Madigan, D., Raftery, A., and Kronmal, R. (1997) Bayesian Model Averaging in Proprtional Hazard Models: Assessing the Risk of a Stroke. Applied Statistics 46: 433-448.

Yeung, K.Y., Bumgarner, R.E. and Raftery, A.E. (2005) Bayesian Model Averaging: Development of an improved multi-class, gene selection and classification tool for microarray data. Bioinformatics 21: 2394-2402.

See Also

iterateBMAsurv.train.predict.assess iterateBMAsurv.train.wrapper, iterateBMAsurv.train, singleGeneCoxph, predictBicSurv, predictiveAssessCategory, trainData, trainSurv, trainCens

Examples

library (BMA)
library(iterativeBMAsurv)
data(trainData)
data(trainSurv)
data(trainCens)

## Perform 1 run of 2-fold cross validation on the training set, using p=10 genes and nbest=5 for fast computation
cv <- crossVal (exset=trainData, survTime=trainSurv, censor=trainCens, diseaseType="DLBCL", noRuns=1, noFolds=2, p=10, nbest=5)

## Upon completion of this function, all relevant output files will be in the working R directory.