R/gbm-data.r

In gbm-developers/gbm3: Generalized Boosted Regression Models

# GBM data object
# Object storing the data used in training a gbm model. 
# 
# @usage gbm_data(x, y, weights, offset)
# 
# @param x matrix or data-frame of the predictor variables 
# 
# @param y matrix or vector of response variables
# 
# @param weights vector containing the weights applied to each data row
# 
# @param offset vector containing the response variables offsets
# 
# @author James Hickey
# 
# @return a gbm_data object 
# 
# @export 
# 
gbm_data <- function(x, y, weights, offset) {
    
    # Check inputs
    verify_data(x, y)
    
    # Convert y if 2-level factor
    y <- convertY(y)
  
    # Check weights and offsets are doubles  
    weights <- check_weights(weights, length(y))
    if((length(weights)==0) || 
       any(is.infinite(weights)) || 
       !is.atomic(weights) ||
       !is.double(weights)) {
      stop("Weights must be a vector of finite doubles")
    }
    
    if((length(offset)<=1) || 
       any(is.infinite(offset)) || 
       !is.atomic(offset) ||
       !is.double(offset)) {
      stop("Offsets must be a vector of finite doubles")
    }
    
    # Store original data - before any formatting
    data <- cbind(as.data.frame(y), as.data.frame(x))
    
    return(structure(list(x=x, y=y, original_data=data, 
                          weights=weights, offset=offset),
                     class="GBMData"))
}