slr {slr}R Documentation

Smoothed logistic regression with Cauchy-Normal random effects

Description

Fit logistic regression model with Cauchy-Normal random probes effects: an application to aCGH data

Usage

slr(y, Z, Chrom=NULL, beta = NULL, b0 = NULL, lambda1 = 1, lambda2=1, maxiter = 15, tol = 1e-10)

Arguments

y binary response variable
Z matrix of predictor variables; rows correspond to probes and columns correspond to subjects.
Chrom vector of chromsone where probes are located; only needed if probes spanned multiple chromosomes
beta starting values for intercept coefficient (optional)
b0 starting values for random-effect (probes) coefficients (optional)
lambda1 smoothing parameter for the Cauchy part; it is recommended that users use cross-validation (cv.lambda) to find the optimal smoothing parameters.
lambda2 smoothing parameter for the Normal part; The default is lambda2=lambda1 which will give a 50-50 mixture between Normal and Cauchy distribution for the random effects.
maxiter maximum iterations allowed.
tol relative convergence critetion.

Value

A list with the following components
beta intercept estimate
b random-effect estimates

Author(s)

Huang Jian and Agus Salim

References

Huang, J., Lei, K., Salim, A., O'Sullivan, K., Pawitan, Y. Classification of array CGH data using smoothed logistic regression model. submitted

See Also

oscc, phenoscc, simacgh, cv.lambda

Examples

# example using oral squamous-cell carcinoma data, chromosome 10 only
data(oscc)
data(phenoscc)
Z <- oscc[oscc$Chrom==10,6:80]
y <- phenoscc$p53[!is.na(phenoscc$p53)]
fit <- slr(y=y,Z=Z,Chrom=oscc$Chrom[oscc$Chrom==10],lambda1=exp(3),lambda2=exp(3))

# example using simulated data from a single chromosome
data(simacgh)
fit2 <- slr(y=simacgh[,1], Z=t(simacgh[,-1]),lambda1=exp(10),lambda2=exp(10))
 

[Package slr version 0.1.9 Index]