ljr (version 1.1-0)

ljrf: Perform forward joinpoint selection algorithm with unlimited upper bound.

Description

This function performs the full forward joinpoint selection algorithm based on the likelihood ratio test statistic. The p-value is determined by a Monte Carlo method.

Usage

ljrf(y,n,tm,X,ofst,R=1000,alpha=.05)

Arguments

y

the vector of Binomial responses.

n

the vector of sizes for the Binomial random variables.

tm

the vector of ordered observation times.

X

a design matrix containing other covariates.

ofst

a vector of known offsets for the logit of the response.

R

number of Monte Carlo simulations.

alpha

significance level of the test.

Value

• pvals
  The estimates of the p-values via simulation.
• Coef
  A table of coefficient estimates.
• Joinpoints
  The estimates of the joinpoint, if it is significant.
• wlik
  The maximum value of the re-weighted log-likelihood.

Details

The re-weighted log-likelihood is the log-likelihood divided by the largest component of n.

References

Czajkowski, M., Gill, R. and Rempala, G. (2007). Model selection in logistic joinpoint regression with applications to analyzing cohort mortality patterns. To appear.

See Also

ljrk,ljrb,ljrf2

Examples

N=20
 m=2
 k=0
 beta=c(0.1,0.1,-0.05)
 gamma=c(0.1,-0.05,0.05)
 ofst=runif(N,-2.5,-1.5)
 x1=round(runif(N,-0.5,9.5))
 x2=round(runif(N,-0.5,9.5))
 X=cbind(x1,x2)
 n=rep(10000,N)
 tm=c(1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10)
 eta=ofst+beta[1]+gamma[1]*tm
 if (m>0)
 for (i in 1:m)
  eta=eta+beta[i+1]*X[,i]
 if (k>0)
  for (i in 1:k)
   eta=eta+gamma[i+1]*pmax(tm-tau[i],0)
 y=rbinom(N,size=n,prob=exp(eta)/(1+exp(eta)))
 temp.ljr=ljrf(y,n,tm,X,ofst,R=1000)