getFit: Fit logit model using cross-validation

In novasmedley/gbmSpm: Find predictive sequential patterns for clinical glioblastoma patients

Description

Fit logit model using cross-validation

Usage

getFit(formula, trainData, index, metric = c("prAUC", "rocAUC"), cv = TRUE,
  lasso = F, llength = NULL, lmax = NULL, seed, lambda = NULL,
  verbose = TRUE)

Arguments

formula	char or formula object
trainData	dataframe, rows are samples to be classified, columns are features including sample ids
index	list of ints, each element is an array, see index
metric	char, see aucs
lasso	logical, whether to use lasso regularization
llength	num, number of lambdas to consider up to lmax
lmax	num, maximum lambda to consider, cannot be NULL if lambda is NULL
seed	int, seed for split
lambda	num, consider one lambda, lmax should be NULL if this has a value
verbose
folds	int, number of folds for cross-validation

Details

Trains logits using glmnet and caret

Examples

# use training partition to create folds for CV
data("features_ratechange_sup0.4g60l2z2") # features and labels for each clinical visit
t <- 'rate'
maxgap <- 60
maxlen <- 2

# format
names <- colnames(feats)
feats <- data.frame(id=row.names(feats),feats)
colnames(feats) <- c('id',names)
feats <- prepLaterality(feats)
feats <- prepLocation(feats)
feats <- removeVisits(feats,
                     maxgap=maxgap,
                     maxlength=maxlen,
                     tType=t,
                     save=F,
                     outDir=NA)
labels <- getClassLabels()
needToRemove <- c('id','iois','eventID', # remove ids
                  labels, # remove labels
                  'IDH1') # not interested

# data partitions
train.ids <- sample(feats$id, size=floor(0.80*nrow(feats)), replace = F) # random
feats <- feats[feats$id %in% train.ids,] #training data
ind <- getTrainingFolds(trainEvents=feats,
                        folds=3,
                        seed=1,
                        verbose=T)
feats <- prepLogitData(data = feats,
                      formula = 'survivalIn60 ~ .',
                      labelName = 'survivalIn60',
                      needToRemove=needToRemove)

# logit
fit <- getFit(formula = 'survivalIn60 ~ .',
              trainData = feats,
              index = ind,
              lasso = TRUE,
              llength = 100,
              lmax = 0.2,
              metric = 'rocAUC',
              seed = 1,
              verbose = T)
              

novasmedley/gbmSpm documentation built on May 17, 2019, 10:39 a.m.