UniDR5.R

In jwilsonwhite/DRcurves: Fit a suite of dose-response curves to data

#' UniDR5
#'
#' 5-parameter unimodal dose-response curve, for MLE estimation. Used as an objective function for mle2
#'
#' @param p 1x6 vector of parameters: g = max or min Y value, h = direction and slope, b1 = exponential rate term (left side), a2 = exponential intercept term, b2 = exponential rate term (right side), and s = standard deviation
#' @param Y concentration values for each observation
#' @param C response value for each observation (not used)
#'
#' @return negative log-likelihood of the model fit
#'
#' @export
#'

uniDR5 = function(p,X,Y){

  g = p[1] # max or min
  h = p[2] # direction (+ or -) and intensity
  #a1 = p[3] # exponential intercept term
  b1 = abs(p[3]) # exponential rate term (rate of increase before the hump)
  a2 = p[4] # exponential intercept term
  b2 = (p[5]) # exponential rate term (rate of decrease after the hump)
  s = exp(p[6]) # standard deviation of response (use exp to avoid negative values)
  P = g+h*(1 + b1*X)/(1+exp(-(a2 + b2*X))) # the equation for the curve
  # Assume a normal distribution for the likelihood (probably OK if data are log transformed), this is defining the actual function that you are fitting to the data

  -sum(dnorm(x=Y,mean=P,sd=s,log=TRUE),na.rm=TRUE)}
parnames(uniDR5) = c('g','h','b1','a2','b2','s')