The package includes functions for fitting hierarchical semiparametric regression model to a large number of parametric test statistics
| Package: | hisemi |
| Type: | Package |
| Version: | 1.1-0 |
| Date: | 2017-07-09 |
| License: | GPL version 2 or newer |
| LazyLoad: | yes |
penLik.EMNewton is the major interface function to be called.
coef.hisemit extracts the estimated parameters.
confint.hisemit returns the confidence intervals.
directSum computes the direct sum of matrices.
EMupdate is the EM algorithm used in penLik.EMNewton.
fitted.hisemit extracts the fitted values.
logLik.hisemit returns the log likelihood.
NRupdate is the Newton-Raphson algorithm used in penLik.EMNewton.
plot.hisemit plots the fitted model.
print.hisemit print summary information.
residuals.hisemit returns the residuals.
scaledTMix.null fits the null model with common pi0.
scaledTMix.psat fits the partially saturated model with free pi0 and common scale factor.
scaledTMix.sat fits the completely saturated model with free pi0 and free scale factor.
vcov.hisemit returns the sandwich variance-covariance matrix.
Long Qu, Dan Nettleton, Jack Dekkers (2012). A hierarchical semiparametric model for incorporating inter-gene relationship information for analysis of genomic data. Biometrics, 68(4):1168-1177
# NOT RUN {
## simulate some fake data
G=100 ## for demonstration only. Normally, G should be much larger
sdncp=1.3
n1=n2=5
df=n1+n2-2
set.seed(54457704)
x=runif(G,1,G)
f=function(x)sin(x*pi/1000)+1
Pi.i=1/(1+exp(f(x)))
Z.i=rbinom(G,1,1-Pi.i)
t0.i=rt(G,df)
ncp.i=rnorm(G,0,sdncp)
t1.i=rt(G,df, ncp.i)
t.i=ifelse(Z.i==0,t0.i,t1.i)
## fit model
(plfit=penLik.EMNewton(t.i, x, df, spar=10^seq(0,8,length=30),plotit=FALSE))
(plfit0=scaledTMix.null(t.i, df))
# }
# NOT RUN {
plot(plfit)
plot(t.i, plfit$lfdr, pch='.')
lines(sort(t.i), plfit0$lfdr[order(t.i)], col=2, lwd=3)
# }
Run the code above in your browser using DataLab