sm_predict: Kernel-weighted predictions from regression models

In ddsjoberg/sjosmooth: Kernel smoothing for time to event data

Description

Calculates kernel-weighted predictions from regression models (i.e. outcomes that can be calculated from the predict function).

Usage

sm_predict(data, method, formula, type, newdata = data,
  method.args = NULL, lambda = 1, kernel = "epanechnikov",
  dist.method = "euclidean", verbose = FALSE)

Arguments

data	data frame
method	function to use
formula	formula
type	type of statistic to smooth (e.g. survival, median survival, etc.)
newdata	new data frame. Default is 'data'. Only requires covariates from the RHS of '~' and the time component from the outcome for some survival estimators.
method.args	List of additional arguments passed on to the modelling function defined by 'method'
lambda	The radius of the kernel for tri-cubic, Epanechnikov, and flat kernels. The standard deviation for the Gaussian kernel
kernel	Specifies the kernel to be used: 'epanechnikov', 'tricube', 'gaussian', and 'flat' are accepted. Default is 'epanechnikov'
dist.method	Specifies the distance measure to be used in the kernel. Default is 'euclidean'. Distance measures accepted by
verbose	Return full set of results as an attribute. Default is 'FALSE'

Examples

sm_predict(
  mtcars,
  method = "glm",
  formula = am ~ mpg,
  method.args = list(family = binomial(link = "logit")),
  type = "response"
)

ddsjoberg/sjosmooth documentation built on May 14, 2019, 5:16 p.m.