staRank-package {staRank} | R Documentation |
Ranking variables based on their stability
Detecting all relevant variables from a data set is challenging, especially when only few samples are available and data is noisy. Stability ranking provides improved variable rankings of increased robustness using resampling or subsampling.
# generate dataset d<-replicate(4,sample(1:10,10,replace=FALSE)) rownames(d)<-letters[1:10] # rank aggregation on the dataset using two base methods aggregRank(d, method='mean') aggregRank(d, method='median') # calculate summary statistic from the data summaryStats(d, method='mean') summaryStats(d, method='RSA') # calculating replicate scores from different summary statistics scores<-getSampleScores(d,'mean',decreasing=FALSE,bootstrap=TRUE) scores<-getSampleScores(d,'mwtest',decreasing=FALSE,bootstrap=TRUE) # perform RSA analysis # get RSA format of data rsaData<-dataFormatRSA(d) # set RSA options opts<-list(LB=min(d),UB=max(d),reverse=FALSE) # run the RSA analysis r<-runRSA(rsaData,opts) # directly obtain the per gene RSA ranking from the data r<-uniqueRSARanking(rsaData,opts) # get stable Ranking, stable setsizes and the Pi matrix for default settings # and stability threshold of 0.9 s<-getStability(d,0.9) # run default stability ranking s<-stabilityRanking(d) # using an accessor function on the RankSummary object stabRank(s) # summarize a RankSummary object summary(s) # generate a rank matrix from a RankSummary object getRankmatrix(s)