predict: Evaluate Basis Functions

In basefun: Infrastructure for Computing with Basis Functions

Evaluate Basis Functions

Description

Evaluate basis functions and compute the function defined by the corresponding basis

Usage

## S3 method for class 'basis'
predict(object, newdata, coef, dim = !is.data.frame(newdata), ...)
## S3 method for class 'cbind_bases'
predict(object, newdata, coef, dim = !is.data.frame(newdata), 
         terms = names(object), ...)
## S3 method for class 'box_bases'
predict(object, newdata, coef, dim = !is.data.frame(newdata), ...)

Arguments

object	a basis or bases object
newdata	a list or data.frame
coef	a vector of coefficients
dim	either a logical indicating that the dimensions shall be obtained from the bases object or an integer vector with the corresponding dimensions (the latter option being very experimental
terms	a character vector defining the elements of a cbind_bases object to be evaluated
...	additional arguments

Details

predict evaluates the basis functions and multiplies them with coef. There is no need to expand multiple variables as predict uses array models (Currie et al, 2006) to compute the corresponding predictions efficiently.

References

Ian D. Currie, Maria Durban, Paul H. C. Eilers, P. H. C. (2006), Generalized Linear Array Models with Applications to Multidimensional Smoothing, Journal of the Royal Statistical Society, Series B: Methodology, 68(2), 259–280.

Examples

  ### set-up a Bernstein polynomial
  xv <- numeric_var("x", support = c(1, pi))
  bb <- Bernstein_basis(xv, order = 3, ui = "increasing")
  ## and treatment contrasts for a factor at three levels
  fb <- as.basis(~ g, data = factor_var("g", levels = LETTERS[1:3]))

  ### join them: we get one intercept and two deviation _functions_
  bfb <- b(bern = bb, f = fb)

  ### generate data + coefficients
  x <- mkgrid(bfb, n = 10)
  cf <- c(1, 2, 2.5, 2.6)
  cf <- c(cf, cf + 1, cf + 2)

  ### evaluate predictions for all combinations in x (a list!)
  predict(bfb, newdata = x, coef = cf)

  ## same but slower
  matrix(predict(bfb, newdata = expand.grid(x), coef = cf), ncol = 3)

basefun documentation built on May 31, 2023, 8:14 p.m.