Single Variable Model in R

single_var_model.R

## How to create a single variable model in R
single_variable_model <- function(x, y, pos) {
  if (class(x) %in% c("numeric", "integer")) {
    # if numeric descretize it
    probs <- unique(quantile(x, probs = seq(0.1, 1, 0.1), na.rm = T))
    x <- cut(x, breaks = probs, include.lowest = T)
  }
  prob_table <- table(as.factor(y), x)
  vals <- unique(y)
  neg <- vals[which(vals != pos)]
  outcome_prob <- sum(y == pos, na.rm = T) / length(y) #outcome probability
  cond_prob <-
    (prob_table[pos,] + 0.001 * outcome_prob) / # 0.001 is laplace smoothing
    (colSums(prob_table) + 0.001) # probability of outcome given variable
  cond_prob_model <- cond_prob[x]
  cond_prob_model[is.na(cond_prob_model)] <- outcome_prob # replace NA values with the most common one
  cond_prob_model
}