mRMRe (version 2.0.3)

mRMRe.Data-class: Class "mRMRe.Data"

Description

mRMRe.Data is the class containing datasets. Most if not all of the routines in the mRMRe package use mRMRe.Data objects as primary input. Such an object is instantiated with a data frame containing the sample sets and optionally, stratum, weight vectors and a prior matrix. In addition to basic accession functions, we describe several methods which serve to manipulate the contents of the dataset. Note that mRMR.data function is a wrapper to easily create mRMRe.Data objects.

Instantiation

Objects are created via calls of the form new("mRMRe.Data", data, strata, weights, priors). data: is expected to be a data frame with samples and features respectively organized as rows and columns. The columns have to be of type :{numeric, ordered factor, Surv} and respectively interpreted as :{continuous, discrete and survival} variables. strata: is expected to be a vector of type :{ordered factor} with the strata associated to the samples provided in data. weights: is expected to be a vector of type :{numeric} with the weights associated to the samples provided in data. priors: is expected to be a matrix of type :{numeric} where priors[i, j]: denotes an forced association between features i and j in data. The latter takes into consideration the directionality of the relationship and must be a value between 0 and 1.

Mutual Information Matrix

The mim method computes and returns a mutual information matrix. A correlation between continuous features is estimated using an estimator specified in continuous_estimator; currently, :{pearson, spearman, kendall, frequency} are supported. The estimator for discrete features is Cramer's V and for all other combinations, concordance index. When outX is set to TRUE, ties are ignored when computing the concordance index and otherwise, these are considered. The correlations are first computed per strata and these are then combined by the inverse variance weight mean of the estimates using a bootstrap_count number of bootstraps if the former parameter is greater than 0, and by the relative weights of each strata otherwise. The resulting correlation is then summated with the corresponding value in the priors matrix with the latter being weighed for a proportion prior_weight of a final, biased correlation.

See Also

mRMRe.Filter-class, mRMRe.Network-class

Examples

showClass("mRMRe.Data")

## load data
data(cgps)

## equivalent ways of building an mRMRe.Data object
ge <- mRMR.data(data = data.frame(cgps.ge[ , 1:10, drop=FALSE]))
ge <- new("mRMRe.Data", data = data.frame(cgps.ge[ , 1:10, drop=FALSE]))

## print data
print(featureData(ge)[1:3, 1:3])

## print feature names
print(featureNames(ge))

## print the first sample names
print(head(sampleNames(ge)))

## print the first sample weights
print(head(sampleWeights(ge)))