R/tik.R

In CorReg: Linear Regression Based on Linear Structure Between Variables

# x est le point etudie
# nbclust le nombre de classes sur xj
# mixmod le tableau mixmod pour xj
tik <- function(x = x, nbclust = nbclust, mixmod = mixmod) {
   prob = rep(0, times = nbclust)
   for (i in 1:nbclust) {
      prob[i] = mixmod[i, 1] * dnorm(x, mean = mixmod[i, 2], sd = sqrt(mixmod[i, 3]))
      if (is.na(prob[i])) {
         message("NA in tik")
         prob[i] = 0
      }
   }
   sumprob = sum(prob)
   if (sumprob == 0) {
      prob[1] = 1
   } else{
      prob = prob / sumprob
   }
   
   return(prob)
}