MEml: Calls the MEml models: MEgbm, MErf, and MEglm. The training...

In nguforche/MEml: Mixed Effect Machine Learning

Description

Calls the MEml models: MEgbm, MErf, and MEglm. The training and test data can be split into lagged training and testing as described in [1]

Usage

MEml_lag(lag = NULL, classifier, dat, id, rhs.vars, resp.vars,
  order.vars, rand.vars = NULL, reg.vars = NULL, part.vars = NULL,
  para, max.iter = 10, seed = 1, return.model = TRUE)

MEml(classifier, dat.trn, dat.tst, id, rhs.vars, resp.vars,
  rand.vars = NULL, reg.vars = NULL, part.vars = NULL, para,
  max.iter = 10, seed = 1, return.model = FALSE, ...)

MEml2(method, data, id, resp.vars, rhs.vars, rand.vars = NULL, para, ...)

Arguments

lag	time lag between predictors and outcome: e.g if lag = 1, then we use predictors in current vist to predict outcome in the next visit.
dat	data frame with predictors and outcome
id	character name of the column containing the group identifier
rhs.vars	caracter vector of predictors
order.vars	order variables (usually time variable)
rand.vars	random effect variables
reg.vars	reg.vars regressors for MOB
part.vars	partitioning variables for MOB and predictors
para	named list of gbm training parameters
max.iter	maximum number of iterations
return.model	should the train model be return. Otherwise the return values is only the performance metrics
method	string name of model. See names(MEml_train()) for list of possible models.

Details

1. MEml_lag Takes the full data set and calls LongiLagSplit to split data into lagged training and testing. MEml_lag also trains the MOB and CTree models (see [1]).

2. MEml is the same as MEml_lag, except that you pass in the training and test set. So you can call LongiLagSlit and pass the derived training and test sets to MEml.

3. MEml2 is the same as MEml, except that you don't pass in the test set. Also, it is currently implemented only for the GLMER, MEgbm, MErf, and GBM models.

Value

The train MEml model and performance matrics (as data frame) if return.model = TRUE

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/MEml documentation built on April 20, 2020, 7:26 a.m.