shipunov (version 1.0)

Dev: Which object is predicted with less accuracy?

Description

Allows to know which object is predicted with less accuracy

Usage

Dev(pred, useNA = "no", adj = FALSE)

Arguments

pred

Pedictions as character matrix, each boot is a column

useNA

Use NAs?

adj

Adjust deviation coefficients? (see below)

Value

Numeric vector of deviation coefficients

Details

Allows to know which object is predicted with less accuracy.

It calculates deviation coeffitient out of the table where rows are objects, and columns are (bootstrapped) predictions.

By default, does not use NA's (use useNA="always" on your risk).

Deviation coefficient is a minimal absolute deviation from one of the range ends (0 and number of predictions), divided by number of predictions, multiplied by number of deviations minus one, and (optionally) adjusted by division to number of predition levels (to make different situations comparable).

See Also

BootRF, BootKNN

Examples

# NOT RUN {
## could be slow
iris.bootrf <- BootRF(iris[, -5], iris[, 5])
data.frame(Iris=make.names(iris[, 5], unique=TRUE), Dev=Dev(iris.bootrf))
# }