R/clogitf.R

In mcneney/logistlogF: Log-F-Penalized Logistic and Conditional Logistic Regression

#' Title
#'
#' @param formula
#' @param data
#' @param pl
#' @param penalty
#' @param maxit
#' @param alpha
#'
#' @return
#' @export
#'
#' @examples
clogitf = function(formula,data,pl=TRUE,penalty=0.5,maxit=50,alpha=0.05) { # Suggestion from Heinze
  require(coxphf)
  data$start = data$matchedset
  data$stop = data$matchedset+0.1
  # Change response as done in clogit()
  newformula = formula
  newformula[[2]]=substitute(Surv(start,stop,case),list(case=formula[[2]]))
  environment(newformula) = environment(formula)
  coxphf(newformula,data,pl,penalty=penalty,maxit=maxit,alpha=alpha)
}

#' Title
#'
#' @param formula
#' @param data
#' @param lkhdDrop
#' @param pl
#' @param penalty
#' @param maxit
#'
#' @return
#' @export
#'
#' @examples
clogitfMidCI <- function(formula,data,lkhdDrop,pl=TRUE,penalty=0.5,maxit=50) {
  alpha <- 1-pchisq(2*lkhdDrop,1)
  if(alpha==1) {
    ff <- clogitf(formula,data,pl=pl,penalty=penalty,maxit=maxit)
    cimid<-ff$coef[1]
  } else {
    ff <- clogitf(formula,data,pl=pl,penalty=penalty,maxit=maxit,alpha=alpha)
    ci <- confint.clogitf(ff)
    cimid <- NA; if(!any(is.infinite(ci))) cimid <- mean(ci)
  }
  return(list(coef = cimid, iter = ff$iter))
}

#' Title
#'
#' @param ff
#'
#' @return
#' @export
#'
#' @examples
confint.clogitf <- function(ff) {
  conf.int = c(ff$ci.lower[1],ff$ci.upper[1])
  conf.int = log(conf.int) #CI from output is for exp(beta)
  if(is.na(conf.int[1])) conf.int[1] = -Inf
  if(is.na(conf.int[2])) conf.int[2] = Inf
  return(conf.int)
}