caret (version 5.07-001)

icr.formula: Independent Component Regression

Description

Fit a linear regression model using independent components

Usage

## S3 method for class 'formula':
icr(formula, data, weights, ..., subset, na.action, contrasts = NULL)
## S3 method for class 'default':
icr(x, y, ...)
## S3 method for class 'icr':
predict(object, newdata, ...)

Value

• For icr, a list with elements
• model
  the results of lm after the ICA transformation
• ica
  pre-processing information
• n.comp
  number of ICA components
• names
  column names of the original data

Details

This produces a model analogous to Principal Components Regression (PCR) but uses Independent Component Analysis (ICA) to produce the scores. The user must specify a value of n.comp to pass to fastICA.

The function preProcess to produce the ICA scores for the original data and for newdata.

See Also

fastICA, preProcess, lm

Examples

data(BloodBrain)

icrFit <- icr(bbbDescr, logBBB, n.comp = 5)

icrFit

predict(icrFit, bbbDescr[1:5,])