DCDI1: Fits dose interval using DC-algorithm

In lixiaomao/personalized-dosing-interval: Determining Personalized Dosing Interval

Description

This function is the wrapper of fitting functions for doseInt

Usage

DCDI1(train, method = "rq", type = "PDI", pred0 = NULL, maxiter = 20,
  step = 1, trace = 0, lower = TRUE, eps = NULL, aL = NULL,
  aU = NULL, lambda = 1e-07, cost = 1, embed.mtry = 1/2,
  svmKernel = NULL, ...)

Arguments

train	the training data set.
method	specified methods used for DC and ordinal approaches.
type	indicates whether PDI or EDI is to be calculated.
pred0	the initial prediction for training data set, if NULL, default initialization is used
maxiter	maximal iterations for DC algorithm
step	stepsize for DC algorithm
trace	the level of detail to be printed: 1,2,3,4
lower	True or False, Whether lower boundary is considered when applied to one-sided interval
nfolds	the number of folds used in cross-validation.
alpha	the guaranteed probability in PDI and the relative loss in EDI. Default is c(0.5,0.5).
two.sided	indicator of whether two-sided interval is considered.
breaks	ways 'as.ordinal' cuts continuous treatment into several ordinal levels, default 'quantile'. User can also choose 'uniform' or specify customized breaks
Eps	parameter for the relaxation
Lambda	penalty for the quantile regression
Cost	cost for the support vector machine
Embed.mtry	proportion of mtry

Examples

  train1=Scenario1.continuous(1000,5,1)
  test1=Scenario1.continuous(10000,5,1)
  fit1=DCDI1(train1,alpha=c(0.5,0.5),method='rq',trace=3)
  predNew1=predict(fit1,test1)
  cor(predNew1$pred,test1$opt)^2

  train3=Scenario3.continuous(2000,5,3)
  test3=Scenario3.continuous(10000,5,3)
  fit3=DCDI2(train3,alpha=c(0.5,0.5),method='rq',trace=3)
  predNew3=predict(fit3,test3)
  cor(predNew3$pred_L,test3$opt_L)^2
  cor(predNew3$pred_R,test3$opt_R)^2

lixiaomao/personalized-dosing-interval documentation built on May 16, 2019, 9:14 p.m.