regclass (version 1.6)

confusion_matrix: Confusion matrix for logistic regression models

Description

This function takes the output of a logistic regression created with glm and returns the confusion matrix.

Usage

confusion_matrix(M,DATA=NA)

Arguments

M

A logistic regression model created with glm

DATA

A data frame on which the confusion matrix will be made. If omitted, the confusion matrix is on the data used in M. If specified, the data frame must have the same column names as the data used to build the model in M.

Details

This function makes classifications on the data used to build a logistic regression model by predicting the "level of interest" (last alphabetically) when the predicted probability exceeds 50%.

See Also

glm

Examples

# NOT RUN {
  #On WINE data as a whole
  data(WINE)
  M <- glm(Quality~.,data=WINE,family=binomial)
  confusion_matrix(M)
  
  #Calculate generalization error using training/holdout
  set.seed(1010)
  train.rows <- sample(nrow(WINE),0.7*nrow(WINE),replace=TRUE)
  TRAIN <- WINE[train.rows,]
  HOLDOUT <- WINE[-train.rows,]
  M <- glm(Quality~.,data=TRAIN,family=binomial)
	confusion_matrix(M,HOLDOUT)
	
	
	#Predicting donation
	#Model predicting from recent average gift amount is significant, but its
	#classifications are the same as the naive model (majority rules)
	data(DONOR)
	M.naive <- glm(Donate~1,data=DONOR,family=binomial)
	confusion_matrix(M.naive)
	M <- glm(Donate~RECENT_AVG_GIFT_AMT,data=DONOR,family=binomial)
	confusion_matrix(M)
	
	 
# }