Description Usage Arguments Value Author(s) References Examples
Z covariates observed only for cases and controls
X observed for all subjects in the study such as sex and age
The distribution of Z given X needs is specified non-parametrically on a number of strata specified by X through the stratum variable.
Intensity model
λ_{X,Z}(t)) = λ_S(X) \exp( β^T Z )
The baseline stratification based on S(X) is specified by baseline.stratum.
When fullobsX is given the baseline may be also on Cox form
λ_{X}(t)) = λ_{S(X_1)} \exp( β_X^T X_2 )
where a stratum is used for some covariates and other covariates are modelled as regression effects. This model is specified by fullobsX=X_2, and S(X_1)= baseline.stratum
When stratZ==1 the baseline is stratified by the Z components as specified by baseline.stratum=S(Z)
λ_{X,Z}(t)) = λ_S(Z) \exp( β^T X )
1 2 3 |
Zobs |
covariates observed for cases and controls, fullobsX is also given it means that the first columns of Zobs is the Z from the above intensity and the second part is the covariates X. |
Tobs |
observed survival times for cases and controls. |
status |
censoring status for observed cases and controls. |
tau |
end of observation period. If tau has length equal to the number of subjects that are not cases and controls it is individual censoring times. |
nno |
number of subjects in the cohort that are not cases or controls. |
Nit |
number of itterations for EM algorithm. |
beta |
starting value for regression parameter. |
detail |
prints out iteration details. |
betait |
number of itterations for Cox regression score in EM, should be at least 2. |
em.dif |
constant used for EM aided differentation, default is (em.dif/number of subjects in cohort). |
stratum |
case specific strata for number of strata for specification of the distribution of Z given X. |
baseline |
starting value for baseline estimates. |
emvar |
computes EM based variance by EM aided differentation. |
fullobsZ |
The fully observed covariates if these are needed for a possible regression model for the baseline |
baseline.stratum |
stratum for the baseline, defined as the grouping given by S(X), a vector of length ntot=nno+nrow(Zobs). |
stratZ |
stratifies after the Z covariate that is only observed for cases and controls, stratification given by baseline.stratum |
returns an object of type "em.ncc". With the following arguments:
cum |
cumulative timevarying regression coefficient estimates are computed within the estimation interval. |
baseline |
the baseline estimates. |
beta |
estimate of parametric components of model. |
var.beta |
variance for beta. |
p |
the distribution of covariates. |
delta |
something about convergence. |
em.dif |
parameter used for EM aided differentation. |
covz |
the covariates realtes do p. |
konv |
something about convergence. |
Thomas Scheike
Scheike and Juul, Biostatistics
Scheike and Maritnussen, SJS
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 | ###library(nccMLE)
# ud<-simNCC(3728,2,rate=1/0.004,win=1) realistic simulation
ud<-simNCC(500,2,rate=1/0.04,win=1)
Zobs<-ud$Xobs; Tobs<-ud$Tobs; status<-ud$status;
nno<-ud$nno; tau<-15;
###################################################################
# Fits model with one common baseline and Z independent of X
# \lambda_0(t) \exp( Z^T \beta)
###################################################################
out<-em.ncc(Zobs,Tobs,status,tau,nno,Nit=100,beta=c(-.2,.2),
betait=20,detail=0,emvar=1)
print(c(out$beta,diag(out$var)^.5)); # estimates related to Z
plot(out$baseline,type="l") # baseline estimates
#### using numDeriv for second derivative
out<-em.ncc(Zobs,Tobs,status,tau,nno,Nit=100,beta=c(-.2,.2),
betait=20,detail=0,emvar=2)
print(c(out$beta,diag(out$var)^.5)); # estimates related to Z
####################################################################
## slow way of fitting the same model, when censoring not the same
####################################################################
out1<-em.ncc(Zobs,Tobs,status,rep(tau,nno),nno,Nit=100,beta=c(-.2,.2),
betait=20,detail=0,emvar=1)
print(c(out1$beta,diag(out1$var)^.5)); # estimates related to Z
lines(out$baseline,lty=2 ) # baseline estimates
#out$p # distribution of Z
###################################################################
# Fits model : \lambda_S(X)(t) \exp( Z^T \beta)
# Z | S(X) is equivalent gives conditional distribution of Z | X
###################################################################
strat<-rbinom(nrow(Zobs),1,0.5)
stratnoobs<-rbinom(500-nrow(Zobs),1,0.5)
out.strat<-em.ncc(Zobs,Tobs,status,tau,nno,Nit=100,beta=0,betait=20,
detail=0,stratum=c(strat,stratnoobs),emvar=1,
baseline.stratum=c(strat,stratnoobs))
print(cbind(out.strat$beta,diag(out.strat$var)^.5));
plot(out.strat$baseline[,1:2],type="l") # baseline estimates
lines(out.strat$baseline[,c(1,3)],type="l") # baseline estimates
out.strat$p # distribution of Z | S(X)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.