ljrb {ljr} | R Documentation |
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.
ljrb(K,y,n,tm,X,ofst,R=1000,alpha=.05)
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. |
The re-weighted log-likelihood is the log-likelihood divided by the largest component of n.
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. |
The authors are Michal Czajkowski, Ryan Gill, and Greg Rempala. The software is maintained by Ryan Gill rsgill01@louisville.edu.
Czajkowski, M., Gill, R. and Rempala, G. (2007). Model selection in logistic joinpoint regression with applications to analyzing cohort mortality patterns. To appear.
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)