ljr (version 1.1-0)

ljrb: Perform backward joinpoint selection algorithm with upper bound K.

Description

This function performs the backward joinpoint selection algorithm with K maximum possible number of joinpoints based on the likelihood ratio test statistic. The p-value is determined by a Monte Carlo method.

Usage

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

Arguments

K

the pre-specified maximum possible number of joinpoints

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,ljrf,ljrb2

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=ljrb(2,y,n,tm,X,ofst,R=1000)