RFreak (version 0.3-0)

robreg.evol: Robust Evolutionary Regression

Description

Carries out robust regression with an evolutionary algorithm. ltsreg, ltareg, lmsreg, lqsreg, and lqdreg are wrappers.

Usage

robreg.evol(x, y, method = c("lts", "lta", "lms", "lqs", "lqd"), 
	quantile=NULL, adjust=FALSE, runs=1, generations=10000, duration=0)
## Wrappers:
ltsreg.evol(...)
ltareg.evol(...)
lmsreg.evol(...)
lqsreg.evol(...)
lqdreg.evol(...)

Arguments

x

Matrix or data frame containing the explanatory variables

y

Vector with the response variables

method

The method to be used. One of "lts", "lta", "lms", "lqs", and "lqd".

quantile

The quantile to be used: see Details.

adjust

Whether to perform intercept adjustment at each step

runs

Number of independent runs

generations

Number of generations after which the algorithm will be stopped

duration

Duration in seconds after which the algorithm will be stopped

...

Arguments to be passed to the default method

Value

The function robreg.evol returns an object of class "evolreg". This object contains:

summary

Summary of the FrEAK run

best

The best subset found

coefficients

Vector of coefficient estimates

crit

The value of the objective function of the regression method

Details

Suppose there are n data points and p regressors, including any intercept.

The first four methods minimize some function of the sorted squared residuals. For methods "lqs" and "lms" it is the quantile squared residual, and for "lts" ("lts") it is the sum of the quantile smallest squared (absolute) residuals. "lqd" minimizes approximately the quartile of the absolute residual differences.

References

O. Morell, T. Bernholt, R. Fried, J. Kunert, and R. Nunkesser (2008). An Evolutionary Algorithm for LTS-Regression: A Comparative Study. Proceedings of COMPSTAT 2008. To Appear.

P. J. Rousseeuw (1984), Least Median of Squares Regression. Journal of the American Statistical Association 79, 871--881.

See Also

"evolreg"

Examples

# load example data
data(stackloss)

# compute different regressions
robreg.evol(stackloss[, 1:3], stackloss[, 4], method= "lts", generations=1000)
lqsreg.evol(stackloss[, 1:3], stackloss[, 4], generations=1000)
lqdreg.evol(stackloss[, 1:3], stackloss[, 4], generations=1000)