Uses likelihood parameter estimation to fit non linear models while attempting several starting values.

.fitNLMwCovariates(
  data,
  nonLinModelQuoted,
  linModelQuoted,
  mllsOuterPrev,
  model = c("CR", "Logistic"),
  maxCover = 1L,
  starts = NULL,
  lower = NULL,
  upper = NULL,
  nbWorkers = 1L
)

Arguments

data

a data.table or data.frame with all covariates and the response variable. Note that incomplete lines are removed.

nonLinModelQuoted

The non-linear equation as a call (quoted expression) passed to bbmle::mle2(minuslog1). See ?mle. Accepts equations with three parameters 'A', 'p' and 'k'.

linModelQuoted

A list of linear equations/modes relating each parameter ('A', 'p' and 'k') with a set of covariates. A call (quoted expression) passed to mle2(..., parameters). Note that for the purpose of tree growth, the linear equation determining 'A' should include a 'cover' predictor indicating the tree cover or dominance in the stand. Should be scaled between 0 and maxCover.

mllsOuterPrev

the output of a previous fitNLMwCovariates run which is used to extract last best AIC and maximum biomass estimate and judge if new iterations are better.

model

character. Non-linear model form used to estimate average maximum biomass. One of "CR" (Chapman-Richards) or "Logistic". In both cases, maximum biomass is equivalent to the 'A' asymptote parameter, which is estimated using observed mean values of predictors entering its linear equation and cover == maxCover, if this predictor is included (as it should). Passed to extractMaxB

maxCover

numeric. Value indicating maximum cover/dominance.

starts

data.table or data.frame of parameter starting values. Will be coerced to named list with names being parameter names.

lower

passed to bbmle::mle2

upper

passed to bbmle::mle2

nbWorkers

integer. If > 1, the number of workers to use in parallelly::makeClusterPSOCK(nbWorkers = .), otherwise no parallellisation is done.

Value

a list with entries mll (the maximum likelihood-estimated coefficients) and AICbest (the AIC of the best models generating these coefficients)

See also