MErf: Mixed Effect random forest

In nguforche/Vira: Virtual Inteligent Robot Assistant

Description

Trains a Mixed Effect random forest for longitudinal continuous, binary and count data. A rule based version or these methods using the inTree package is also implemented(see [1])

Usage

MErfRules(form, dat, groups = NULL, rand.vars = "1", para = NULL,
  tol = 0.00001, max.iter = 100, include.RE = FALSE,
  verbose = FALSE, maxdepth = 5,
  glmer.Control = glmerControl(optimizer = "bobyqa", check.nobs.vs.nRE =
  "ignore", check.nobs.vs.nlev = "ignore"), nAGQ = 0,
  likelihoodCheck = TRUE, K = 3, decay = 0.05, ...)

MErfRules(form, dat, groups = NULL, rand.vars = "1", para = NULL,
  tol = 0.00001, max.iter = 100, include.RE = FALSE,
  verbose = FALSE, maxdepth = 5,
  glmer.Control = glmerControl(optimizer = "bobyqa", check.nobs.vs.nRE =
  "ignore", check.nobs.vs.nlev = "ignore"), nAGQ = 0,
  likelihoodCheck = TRUE, K = 3, decay = 0.05, ...)

Arguments

groups	character name of the column containing the group identifier
rand.vars	random effect variables
para	named list of gbm training parameters
tol	convergence tolerance
max.iter	maximum number of iterations
include.RE	(logical) to include random effect Zb as predictor in gbm?
verbose	verbose for lme4
likelihoodCheck	(logical) to use log likelihood of glmer to check for convergence?
...	Further arguments passed to or from other methods.
X	data.frame with predictors
Y	binary response vector
lme.family	glmer control
type	of predictions of gbm to pass to lme4 as population estimates (these will be used as offset)

Value

An object of class MEgbm; a list with items

rffit	fitted random forest model
glmer.fit	fitted mixed effect logistic regression model
logLik	log likelihood of mixed effect logistic regression
random.effects	random effect parameter estimates
glmer.form	lmer4 formula
glmer.CI	estimates of mixed effect logistic regression with approximate confidence intervals on the logit scale. More accurate values can be obtained by bootstrap
fitted.probs	fitted probabilites for final model
fitted.class	fitted class labels for final model
train.perf	various performance measures for final model on training set
threshold	classification cut-off

Author(s)

Che Ngufor <[e-mail]>

References

Che Ngufor, Holly Van Houten, Brian S. Caffo , Nilay D. Shah, Rozalina G. McCoy Mixed Effect Machine Learning: a framework for predicting longitudinal change in hemoglobin A1c, in Journal of Biomedical Informatics, 2018

nguforche/Vira documentation built on June 2, 2019, 9:08 p.m.