econum: Various Tools Developed at EcoNum Lab

# Calibration of chemical determination curves and prediction of unknowns
# OO-oriented version of the functions in the chemCal package
# Needs the chemCal and MASS packages

#chemCal <- function(conc, resp, data, intercept = TRUE, robust = FALSE,
#weights = NULL, ...) {
#  cl <- match.call()
#  if (missing(data)) {
#    if (isTRUE(intercept)) form <- resp ~ conc else form <- resp ~ conc + 0
#    if (isTRUE(robust)) {# Use robust rlm() function from MASS
#      obj <- rlm(form, weights = weights, ...)
#    } else { # Use classical lm()
#      obj <- lm(form, weights = weights, ...)
#    }
#  } else {# data is provided, conc and resp are the name of the variables
#    conc <- sub('^"(.*)"$', "\\1", deparse(substitute(conc)))
#    resp <- sub('^"(.*)"$', "\\1", deparse(substitute(resp)))
#    form <- paste(resp, "~", conc)
#    if (!isTRUE(intercept)) form <- paste(form, "+ 0")
#    form <- as.formula(form)
#    if (isTRUE(robust)) {# Use robust rlm() function from MASS
#      obj <- rlm(form, data = data, weights = weights, ...)
#    } else { # Use classical lm()
#      obj <- lm(form, data = data, weights = weights, ...)
#    }
#  }
#  # Change the call element of the object
#  obj$call <- cl
#  # Return a 'chemCal' object
#  class(obj) <- c("chemCal", class(obj))
#  obj
#}
#
## print(), summary(), coef() methods are those of lm() or rlm() functions
## plot() and predict() are specific
## lod() and loq() methods are defined in the chemCal package
#plot.chemCal <- function(x, which = 0L, ...) {
#  if (which == 0L) { # Use calplot()
#    calplot(x, ...)
#  } else {# Use the regular plot.lm() function
#    # Not possible to mix both
#    if (0 %in% which)
#      stop("Cannot mix calibration plot (which == 0) and residual analysis graphs (which > 0)")
#    plot.lm(x, which = which, ...)
#  }
#}
#
#predict.chemCal <- function(object, newdata, samples, ..., ws = "auto",
#alpha = 0.05, var.s = "auto") {
#  # On the contrary to inverse.predict, this function allows to perform the
#  # calculation for many samples at a time
#  ## TODO...
#}