| predict,biosign-method {biosigner} | R Documentation |
This function predicts values based upon biosign classifiers trained
on the 'S' signature
## S4 method for signature 'biosign' predict(object, newdata, tierMaxC = "S", ...)
object |
An S4 object of class |
newdata |
Either a data frame or a matrix, containing numeric columns only, with column names identical to the 'x' used for model training with 'biosign'. |
tierMaxC |
Character: Maximum level of tiers to display: Either 'S'or 'A'. |
... |
Currently not used. |
Data frame with the predictions for each classifier as factor columns.
Philippe Rinaudo and Etienne Thevenot (CEA)
## loading the diaplasma dataset
data(diaplasma)
attach(diaplasma)
## restricting to a smaller dataset for this example
featureSelVl <- variableMetadata[, "mzmed"] >= 490 & variableMetadata[, "mzmed"] < 500
dataMatrix <- dataMatrix[, featureSelVl]
variableMetadata <- variableMetadata[featureSelVl, ]
## training the classifiers
## a bootI = 5 number of bootstraps is used for this example
## we recommend to keep the default bootI = 50 value for your analyzes
set.seed(123)
diaSign <- biosign(dataMatrix, sampleMetadata[, "type"], bootI = 5)
## fitted values (for the subsets restricted to the 'S' signatures)
sFitDF <- predict(diaSign)
## confusion tables
print(lapply(sFitDF, function(predFc) table(actual = sampleMetadata[,
"type"], predicted = predFc)))
## balanced accuracies
sapply(sFitDF, function(predFc) { conf <- table(sampleMetadata[,
"type"], predFc)
conf <- sweep(conf, 1, rowSums(conf), "/")
mean(diag(conf))
})
## note that these values are slightly different from the accuracies
## returned by biosign because the latter are computed by using the
## resampling scheme selected by the bootI or crossvalI arguments
getAccuracyMN(diaSign)["S", ]
detach(diaplasma)