twilight.getlambda {twilight}R Documentation

Internal function to compute regularization parameter lambda

Description

The SEP method optionally includes a penalty term in order to estimate the percentage of non-significant differences/genes.

Usage

twilight.getlambda(xin, verbose = TRUE)

Arguments

xin A numerical vector containing p-values.
verbose Logical value for message printing.

Details

The function finds a suitable regularization parameter λ. For a sequence of lambdas, the objective function of SEP is computed for subsamples of the p-value vector xin. The final estimate is chosen based upon a Wilcoxon test comparison between objective function values of λ=0 and each λ>0. The penalized objective function should not differ a lot from the unpenalized one. Therefore, the highest lambda that leads to a non-significant difference in means is chosen.

Value

Returns numerical value.

Author(s)

Stefanie Scheid http://www.molgen.mpg.de/~scheid

References

Scheid S and Spang R (2004): A stochastic downhill search algorithm for estimating the local false discovery rate, IEEE TCBB 1(3), 98–108.

See Also

twilight


[Package twilight version 1.4.0 Index]