random forest feature selection based on binomial exact test

Description

cv.binomialRF is the cross-validated form of the binomialRF, where K-fold crossvalidation is conducted to assess the feature's significance. Using the cvFolds=K parameter, will result in a K-fold cross-validation where the data is 'chunked' into K-equally sized groups and then the averaged result is returned.

Usage

.cv_binomialRF(X, y, cvFolds = 5, fdr.threshold = 0.05,
  fdr.method = "BY", ntrees = 2000, keep.both = FALSE)

Arguments

Value

a data.frame with 4 columns: Feature Name, cross-validated average for Frequency Selected, CV Median (Probability of Selecting it randomly), CV Median(Adjusted P-value based on fdr.method), and averaged number of times selected as signficant.

References

Zaim, SZ; Kenost, C.; Lussier, YA; Zhang, HH. binomialRF: Scalable Feature Selection and Screening for Random Forests to Identify Biomarkers and Their Interactions, bioRxiv, 2019.

Examples

set.seed(324)

###############################
### Generate simulation data
###############################

X = matrix(rnorm(1000), ncol=10)
trueBeta= c(rep(10,5), rep(0,5))
z = 1 + X %*% trueBeta
pr = 1/(1+exp(-z))
y = as.factor(rbinom(100,1,pr))

###############################
### Run cross-validation
###############################

random forest feature selection based on binomial exact test

Description

binomialRF is the R implementation of the feature selection algorithm by (Zaim 2019)

Usage

binomialRF(X,y, fdr.threshold = .05,fdr.method = 'BY',
                      ntrees = 2000, percent_features = .5,
                      keep.both=FALSE, user_cbinom_dist=NULL,
                      sampsize=round(nrow(X)*.63))

Arguments

Value

a data.frame with 4 columns: Feature Name, Frequency Selected, Probability of Selecting it randomly, Adjusted P-value based on fdr.method

References

Zaim, SZ; Kenost, C.; Lussier, YA; Zhang, HH. binomialRF: Scalable Feature Selection and Screening for Random Forests to Identify Biomarkers and Their Interactions, bioRxiv, 2019.

Examples

set.seed(324)

###############################
### Generate simulation data
###############################

X = matrix(rnorm(1000), ncol=10)
trueBeta= c(rep(10,5), rep(0,5))
z = 1 + X %*% trueBeta
pr = 1/(1+exp(-z))
y = as.factor(rbinom(100,1,pr))

###############################
### Run binomialRF
###############################
require(correlbinom)

rho = 0.33
ntrees = 250
cbinom = correlbinom(rho, successprob =  calculateBinomialP(10, .5), trials = ntrees, 
                               precision = 1024, model = 'kuk')

binom.rf <-binomialRF(X,y, fdr.threshold = .05,fdr.method = 'BY',
                      ntrees = ntrees,percent_features = .5,
                      keep.both=FALSE, user_cbinom_dist=cbinom,
                      sampsize=round(nrow(X)*rho))

print(binom.rf)

calculate the probability, p, to conduct a binomial exact test

Description

calculateBinomialP returns a probability of randomly selecting a feature as the root node in a decision tree. This is a generic function that is called internally in binomialRF but that may also be called directly if needed. The arguments ... should be, L= Total number of features in X, and percent_features= what percent of L is subsampled in the randomForest call.

Usage

calculateBinomialP(L, percent_features)

Arguments

Value

If L is an integeter returns a probability value for selecting predictor Xj randomly

Examples

calculateBinomialP(110, .4)
calculateBinomialP(13200, .5)

calculate the probability, p, to conduct a binomial exact test

Description

calculateBinomialP_Interaction returns a probability of randomly selecting a feature as the root node in a decision tree. This is a generic function that is called internally in binomialRF but that may also be called directly if needed. The arguments ... should be, L= Total number of features in X, and percent_features= what percent of L is subsampled in the randomForest call.

Usage

calculateBinomialP_Interaction(L, percent_features, K = 2)

Arguments

Value

If L is an integeter returns a probability value for selecting predictor Xj randomly

Examples

calculateBinomialP_Interaction(110, .4,2 )

random forest feature selection based on binomial exact test

Description

binomialRF is the R implementation of the feature selection algorithm by (Zaim 2019)

Usage

geneset_binomialRF(binomialRF_object, gene_ontology, cutoff = 0.2)

Arguments

Value

a data.frame with 4 columns: Geneset Name, P-value, Adjusted P-value based on fdr.method

References

Zaim, SZ; Kenost, C.; Lussier, YA; Zhang, HH. binomialRF: Scalable Feature Selection and Screening for Random Forests to Identify Biomarkers and Their Interactions, bioRxiv, 2019.

random forest feature selection based on binomial exact test

Description

k_binomialRF is the R implementation of the interaction feature selection algorithm by (Zaim 2019). k_binomialRF extends the binomialRF algorithm by searching for k-way interactions.

Usage

k_binomialRF(X, y, fdr.threshold = 0.05, fdr.method = "BY",
  ntrees = 2000, percent_features = 0.3, K = 2, cbinom_dist = NULL,
  sampsize = nrow(X) * 0.4)

Arguments

Value

a data.frame with 4 columns: Feature Name, Frequency Selected, Probability of Selecting it randomly, Adjusted P-value based on fdr.method

References

Zaim, SZ; Kenost, C.; Lussier, YA; Zhang, HH. binomialRF: Scalable Feature Selection and Screening for Random Forests to Identify Biomarkers and Their Interactions, bioRxiv, 2019.

Examples

set.seed(324)

###############################
### Generate simulation data
###############################

X = matrix(rnorm(1000), ncol=10)
trueBeta= c(rep(10,5), rep(0,5))
z = 1 + X %*% trueBeta
pr = 1/(1+exp(-z))
y = rbinom(100,1,pr)

###############################
### Run interaction model
###############################

require(correlbinom)

rho = 0.33
ntrees = 250
cbinom = correlbinom(rho, successprob =  calculateBinomialP_Interaction(10, .5,2), 
                               trials = ntrees, precision = 1024, model = 'kuk')

k.binom.rf <-k_binomialRF(X,y, fdr.threshold = .05,fdr.method = 'BY',
                      ntrees = ntrees,percent_features = .5,
                      cbinom_dist=cbinom,
                      sampsize=round(nrow(X)*rho))

A prebuilt distribution for correlated binary data

Description

This data contains probability mass functions (pmf's) for correlated binary data for various parameters. The sum of correlated exchangeable binary data is a generalization of the binomial distribution that deals with correlated trials. The correlation in decision trees occurs as the subsampling and bootstrapping step in random forests touch the same data, creating a co-dependency. This data contains some pre-calculated distributions for random forests with 500, 1000, and 2000 trees with 10, 100, and 1000 features. For more distributions, they can be calculated via the correlbinom R package.

Usage

pmf_list

Format

A list of lists

References

Witt, Gary. "A Simple Distribution for the Sum of Correlated, Exchangeable Binary Data." Communications in Statistics-Theory and Methods 43, no. 20 (2014): 4265-4280.