clogitRV: Conditional Logistic Regression with Robust Variance

In sampsonj74/clogitRV:

Description

Performs conditional logistic regression with a robust variance estimator

Usage

clogitRV(formula, id.set, data)

Arguments

formula	Similar to a formula for glm.
id.set	The ID variable for matched sets.
data	The dataset

Value

coef The coefficeints (i.e. betas) and their standard error

cov The covariance matrix of the coefficeints (useful for wald tests)

Examples

###A simple Example
n    <-  10
Y    <- rep(c(1,0,1,0),n)
ID   <- rep(c(1:n),each=4)
X1   <- rnorm(4*n)
X2   <- rnorm(4*n)
data  <- cbind(Y,X1,X2,ID)
clogitRV(Y~X1+X2,~ID,data)
##
##
##
##An example with a time-varying coefficient
##simData is an example dataset included in the package
?simData

###Run the Model (note: z is just a covariate)
simData$x1 <- simData$t   * simData$x
simData$x2 <- simData$t^2 * simData$x
model      <- clogitRV(Y~x+x1+x2+z,~ID, data=simData)
beta       <- model$coef[-4,1]
cov        <- model$cov[-4,-4]

##Draw the picture
library(ggplot2)
t2       <- seq(min(simData$t),max(simData$t),length.out=100)
t2M      <- cbind(1,t2,t2^2)
beta2     <- t2M %*% beta
se2       <- sqrt(diag(t2M%*%cov%*%t(t2M)))
plotData  <- as.data.frame(t2,beta2,se2)
ggplot(data=plotData,aes(t2,beta2)) +
  geom_line() +
  geom_point()+
  geom_errorbar(aes(ymin=beta2-1.96*se2, ymax=beta2+1.96*se2))+
  xlab("Time Until Diaganosis")+
  ylab("log(RR)")

##Get the p-values
##P-value for overall effect of the biomarker
1-pchisq(t(beta)%*%solve(cov)%*%beta,length(beta))
##P-value for non-linear effect of the biomarker
1-pchisq(t(beta[-1])%*%solve(cov[-1,-1])%*%beta[-1],length(beta[-1]))

sampsonj74/clogitRV documentation built on Dec. 22, 2021, 10:11 p.m.