cv.summary.bas: Summaries for Out of Sample Prediction

In BAS: Bayesian Variable Selection and Model Averaging using Bayesian Adaptive Sampling

Summaries for Out of Sample Prediction

Description

Compute average prediction error from out of sample predictions

Usage

cv.summary.bas(pred, ytrue, score = "squared-error")

Arguments

pred	fitted or predicted value from the output from predict.bas
ytrue	vector of left out response values
score	function used to summarize error rate. Either "squared-error", or "miss-class"

Value

For squared error, the average prediction error for the Bayesian estimator error = sqrt(sum(ytrue - yhat)^2/npred) while for binary data the misclassification rate is more appropriate.

Author(s)

Merlise Clyde [e-mail]

See Also

predict.bas

Examples

## Not run: 
library(foreign)
cognitive <- read.dta("https://www.stat.columbia.edu/~gelman/arm/examples/child.iq/kidiq.dta")
cognitive$mom_work <- as.numeric(cognitive$mom_work > 1)
cognitive$mom_hs <- as.numeric(cognitive$mom_hs > 0)
colnames(cognitive) <- c("kid_score", "hs", "iq", "work", "age")

set.seed(42)
n <- nrow(cognitive)
test <- sample(1:n, size = round(.20 * n), replace = FALSE)
testdata <- cognitive[test, ]
traindata <- cognitive[-test, ]
cog_train <- bas.lm(kid_score ~ ., prior = "BIC", modelprior = uniform(), data = traindata)
yhat <- predict(cog_train, newdata = testdata, estimator = "BMA", se = F)
cv.summary.bas(yhat$fit, testdata$kid_score)

## End(Not run)

BAS documentation built on Sept. 30, 2024, 9:19 a.m.