lrp: Longitudinal Recursive Partitioning

In Rjacobucci/longRPart2: Recursive Partitioning of Longitudinal Data

Longitudinal Recursive Partitioning

Description

Longitudinal Recursive Partitioning

Usage

lrp(
  method,
  nlme.model = NULL,
  randomFormula,
  fixedFormula = NULL,
  data,
  start,
  group,
  rPartFormula,
  weight = NULL,
  use_parallel = FALSE,
  R = NULL,
  min.dev = NULL,
  control = rpart.control()
)

Arguments

method	Whether to use lme() or nlme(). Use either method="lme" or method="nlme". This changes what additional arguments need to be passed.
nlme.model	Necessary to specify if method="nlme"
randomFormula	Random effects to include for nlme() or lme()
fixedFormula	Fixed effects to include for nlme() or lme()
data	Dataset
start	Starting values for nlme()
group	Grouping for nlme()
rPartFormula	Not sure yet
weight	Sample weights to be passed to rpart
use_parallel	Whether to parallelize the split models
R	Correlation matrix to use for nlme. this is correlation=
min.dev	The minimum decrease in deviance to choose a split. Note that this overrides the default cp criterion in rpart.control()
control	Control function to be passed to rpart()

Examples

library(longRPart2)


data(ex.data.3)
model.0 = nlme(y~b0i+b1i*time,
               data=ex.data.3,
               fixed=b0i+b1i~1,
               random=b0i+b1i~1,
               group=~id,
               start=c(10,5))


lcart.mod1 <- lrp(method="nlme",
                 nlme.model=y~b0i+b1i*time,
                 fixedFormula=b0i+b1i~1,
                 rPartFormula = ~ z,
                 group= ~ id,
                 randomFormula=b0i+b1i~1,
                 data=ex.data.3,
                 start=c(10,5))

data(lcart.mod1)
summary(lcart.mod1)
plot(lcart.mod1)
# for smooth_method, "loess" is recommend but "gam" faster
lrp2Plot(lcart.mod1,smooth_method="gam")

Rjacobucci/longRPart2 documentation built on Nov. 24, 2022, 2:32 a.m.