nlme.mmkin: Create an nlme model for an mmkin row object

Description

This functions sets up a nonlinear mixed effects model for an mmkin row object. An mmkin row object is essentially a list of mkinfit objects that have been obtained by fitting the same model to a list of datasets.

Usage

# S3 method for mmkin
nlme(
  model,
  data = "auto",
  fixed = lapply(as.list(names(mean_degparms(model))), function(el) eval(parse(text =
    paste(el, 1, sep = "~")))),
  random = pdDiag(fixed),
  groups,
  start = mean_degparms(model, random = TRUE, test_log_parms = TRUE),
  correlation = NULL,
  weights = NULL,
  subset,
  method = c("ML", "REML"),
  na.action = na.fail,
  naPattern,
  control = list(),
  verbose = FALSE
)
# S3 method for nlme.mmkin
print(x, digits = max(3, getOption("digits") - 3), ...)
# S3 method for nlme.mmkin
update(object, ...)

Value

Upon success, a fitted 'nlme.mmkin' object, which is an nlme object with additional elements. It also inherits from 'mixed.mmkin'.

Arguments

model: An mmkin row object.
data: Ignored, data are taken from the mmkin model
fixed: Ignored, all degradation parameters fitted in the mmkin model are used as fixed parameters
random: If not specified, no correlations between random effects are set up for the optimised degradation model parameters. This is achieved by using the nlme::pdDiag method.
groups: See the documentation of nlme
start: If not specified, mean values of the fitted degradation parameters taken from the mmkin object are used
correlation: See the documentation of nlme
weights: passed to nlme
subset: passed to nlme
method: passed to nlme
na.action: passed to nlme
naPattern: passed to nlme
control: passed to nlme
verbose: passed to nlme
x: An nlme.mmkin object to print
digits: Number of digits to use for printing
...: Update specifications passed to update.nlme
object: An nlme.mmkin object to update

Details

Note that the convergence of the nlme algorithms depends on the quality of the data. In degradation kinetics, we often only have few datasets (e.g. data for few soils) and complicated degradation models, which may make it impossible to obtain convergence with nlme.

Examples

Run this code

ds <- lapply(experimental_data_for_UBA_2019[6:10],
 function(x) subset(x$data[c("name", "time", "value")], name == "parent"))

if (FALSE) {
  f <- mmkin(c("SFO", "DFOP"), ds, quiet = TRUE, cores = 1)
  library(nlme)
  f_nlme_sfo <- nlme(f["SFO", ])
  f_nlme_dfop <- nlme(f["DFOP", ])
  anova(f_nlme_sfo, f_nlme_dfop)
  print(f_nlme_dfop)
  plot(f_nlme_dfop)
  endpoints(f_nlme_dfop)

  ds_2 <- lapply(experimental_data_for_UBA_2019[6:10],
   function(x) x$data[c("name", "time", "value")])
  m_sfo_sfo <- mkinmod(parent = mkinsub("SFO", "A1"),
    A1 = mkinsub("SFO"), use_of_ff = "min", quiet = TRUE)
  m_sfo_sfo_ff <- mkinmod(parent = mkinsub("SFO", "A1"),
    A1 = mkinsub("SFO"), use_of_ff = "max", quiet = TRUE)
  m_dfop_sfo <- mkinmod(parent = mkinsub("DFOP", "A1"),
    A1 = mkinsub("SFO"), quiet = TRUE)

  f_2 <- mmkin(list("SFO-SFO" = m_sfo_sfo,
   "SFO-SFO-ff" = m_sfo_sfo_ff,
   "DFOP-SFO" = m_dfop_sfo),
    ds_2, quiet = TRUE)

  f_nlme_sfo_sfo <- nlme(f_2["SFO-SFO", ])
  plot(f_nlme_sfo_sfo)

  # With formation fractions this does not coverge with defaults
  # f_nlme_sfo_sfo_ff <- nlme(f_2["SFO-SFO-ff", ])
  #plot(f_nlme_sfo_sfo_ff)

  # For the following, we need to increase pnlsMaxIter and the tolerance
  # to get convergence
  f_nlme_dfop_sfo <- nlme(f_2["DFOP-SFO", ],
    control = list(pnlsMaxIter = 120, tolerance = 5e-4))

  plot(f_nlme_dfop_sfo)

  anova(f_nlme_dfop_sfo, f_nlme_sfo_sfo)

  endpoints(f_nlme_sfo_sfo)
  endpoints(f_nlme_dfop_sfo)

  if (length(findFunction("varConstProp")) > 0) { # tc error model for nlme available
    # Attempts to fit metabolite kinetics with the tc error model are possible,
    # but need tweeking of control values and sometimes do not converge

    f_tc <- mmkin(c("SFO", "DFOP"), ds, quiet = TRUE, error_model = "tc")
    f_nlme_sfo_tc <- nlme(f_tc["SFO", ])
    f_nlme_dfop_tc <- nlme(f_tc["DFOP", ])
    AIC(f_nlme_sfo, f_nlme_sfo_tc, f_nlme_dfop, f_nlme_dfop_tc)
    print(f_nlme_dfop_tc)
  }

  f_2_obs <- update(f_2, error_model = "obs")
  f_nlme_sfo_sfo_obs <- nlme(f_2_obs["SFO-SFO", ])
  print(f_nlme_sfo_sfo_obs)
  f_nlme_dfop_sfo_obs <- nlme(f_2_obs["DFOP-SFO", ],
    control = list(pnlsMaxIter = 120, tolerance = 5e-4))

  f_2_tc <- update(f_2, error_model = "tc")
  # f_nlme_sfo_sfo_tc <- nlme(f_2_tc["SFO-SFO", ]) # No convergence with 50 iterations
  # f_nlme_dfop_sfo_tc <- nlme(f_2_tc["DFOP-SFO", ],
  #  control = list(pnlsMaxIter = 120, tolerance = 5e-4)) # Error in X[, fmap[[nm]]] <- gradnm

  anova(f_nlme_dfop_sfo, f_nlme_dfop_sfo_obs)

}