Code for adaptively spacing knots for a given covariate.

Usage

return.reg.spline.fit(
  response,
  explanatory,
  degree,
  minKnots,
  maxKnots,
  startKnots,
  gap,
  winHalfWidth,
  fitnessMeasure = "BIC",
  maxIterations = 100,
  initialise = TRUE,
  initialKnots = NULL,
  baseModel = NULL,
  bd,
  spl,
  interactionTerm = interactionTerm,
  cv.opts,
  splineParams,
  printout
)

Arguments

response: vector of response data for the modelling process
explanatory: vector of covariate to find knots for
degree: degree of the spline to be used
minKnots: minimum number of knots to fit
maxKnots: maximum number of knots to fit
startKnots: number of equally spaced knots to start with (between minKnots and maxKnots)
gap: minimum gap between knots (in unit of measurement of explanatory)
winHalfWidth: Half-width of window used to calculate region with biggest average residual magnitude
fitnessMeasure: (default=BIC). Measure used to evaluate the fit. Other options are AIC, AICc, BIC, QAIC, QAICc, QBIC (Quasi-Likelihood Information Criterion with log(n) penalty), k-fold cross-validation (cv.gamMRSea), AICtweedie, BICtweedie.
maxIterations: exchange/improve heuristic will terminate after maxIterations if still running
initialise: (default = TRUE). Logical stating whether or not to start with equally spaced knots (TRUE) or user specified locations (FALSE)
initialKnots: If initialise=FALSE then the start locations for the knots are specified in initialKnots
baseModel: starting model for SALSA to use. Must not contain the covariate in explanatory
bd: the x-coordinate of the boundary knots of explanatory
spl: "bs" uses b-spline, "cc" uses cyclic cubic, "ns" uses natural cubic spline for fitting smooth to explanatory
interactionTerm: character stating the variable to use as an interaction
cv.opts: A list object containing options for cv.gamMRSea.
printout: A logical stating whether analysis information is printed in the workspace or into a logfile (TRUE) or not at all (FALSE)
suppress.printout: default=FALSE. If TRUE, progress is printed into the workspace. If FALSE, a .log file is created in the working directory.

Author

Cameron Walker, Department of Engineering Science, University of Auckland, University of Auckland), Lindesay Scott-Hayward (University of St Andrews)