Take an R lm model expressed in terms of y ~ x. Provided values for y, return "estimated" x.

returnXinY.R

returnXinY <- function(lm_model, y_values){
  # This function takes a lm model expressed in terms of y ~ x, and values for y, and returns x
  # Given a model of the form y ~ beta.x + alpha, then given values for y, return x = (y-alpha)/beta
  # Crazy I know, but there you have it.
  #
  if(class(lm_model) != "lm"){
    stop("This function requires a lm mode with a single predictor variable (y ~x)")
  }
  if(length(unlist(strsplit(toString(lm_model$call[2][[1]]), " "))) > 3){
    stop("This function only works with a lm model with a single predictor variable (y ~x)")
  }
  lm_coefs <- coef(lm_model)
  x_values <- (y_values-lm_coefs[1])/lm_coefs[2]
  return(x_values)
}