aggrmodel: Fit Aggregated Model
In gabrielfranco89/aggrmodel: A R package to fit aggregated data model

Description Usage Arguments Value Examples

Fit Aggregated Model

aggrmodel(
  formula = NULL,
  data,
  market,
  Y = NULL,
  timeVar,
  timeVar2 = NULL,
  groupVar,
  repVar,
  n_basis,
  n_basis_cov = NULL,
  basisFunction = "B-Splines",
  n_basis2 = NULL,
  n_order2 = NULL,
  basisFunction2 = NULL,
  cicleRep = FALSE,
  n_order = 4,
  covType = "Homog_Uniform",
  corType = "exponential",
  optSampleCovMatrix = FALSE,
  sigPar_init = NULL,
  corPar_init = NULL,
  tauPar_init = NULL,
  betaCov_init = NULL,
  returnFitted = TRUE,
  positive_restriction = FALSE,
  optimMethod = "L-BFGS-B",
  use_parallel = FALSE,
  n_cores = parallel::detectCores() - 1,
  truncateDec = NULL,
  verbose = FALSE,
  optVerbose = FALSE,
  useGrad = FALSE,
  diffTol = 1e-05,
  itMax = 100
)

`formula`	building...
`data`	Dataset containing group, replicates (if any), time and aggregated signal
`market`	Market data frame. MUST be a 3 column dataframe with the following order: Group, Type and Number of subjects
`Y`	Dependent variable: aggregated signal
`timeVar`	Name of time variable
`timeVar2`	Name of second functional
`groupVar`	Name of grouping variable
`repVar`	Name of replicates variable
`n_basis`	Number of basis functions for basis expansion
`n_basis_cov`	Number of basis functions for variance functional expansion
`basisFunction`	Character indicating which basis: 'B-Splines' or 'Fourier'
`n_basis2`	Number of basis for second functional
`n_order2`	Order for second functional expansion
`basisFunction2`	Character indicating which basis: 'B-Splines' or 'Fourier'
`cicleRep`	Indicator TRUE/FALSE if replicates are cyclical
`n_order`	Order of basis Splines (Default: 4)
`covType`	Covariance functional type. One of "Homog_Uniform", "Homog" or "Heterog"
`corType`	Correlation structure type. One of "periodic" or "exponential"
`optSampleCovMatrix`	Optmization criterion via sample covariance matrix convergence (TRUE and default) or via likelihood (more sensitive)
`sigPar_init`	Inital values for sigma
`corPar_init`	Numeric: Initial value for correlation parameters (default:20)
`tauPar_init`	Numeric: Initial value for expoent parameters of complete covariance (default:0.5)
`betaCov_init`	Inital values for variance functional expansion
`returnFitted`	Should the fitted values be returned in output?
`positive_restriction`	TRUE/FALSE if mean curves are strictly positive
`optimMethod`	Choose optim method (Default: L-BFGS-B)
`use_parallel`	TRUE/FALSE if computation should be parallel
`n_cores`	Number of clusters. Default: parallel::detectCores()
`truncateDec`	Decimal to be truncated at covariance matrix
`verbose`	TRUE/FALSE prints likelihood values during optimization
`optVerbose`	Print parameters while in optmization
`useGrad`	Use gradient function approximation for optimization? (Default: FALSE)
`diffTol`	Tolerance of model covergence
`itMax`	Maximum number of iterations (Default: 100)

An aggrmodel object

set.seed(81453)
df <- createSimuData(B1 = 8,
                    nRep=10)

mkt <- attr(df,"market")
df <- subset(df, group <= 8) # get only cluster 1
mkt <- subset(mkt,group<=8)

fit <-
 aggrmodel(
   data = df,
   market = mkt,
   Y = obs,
   timeVar = time,
   groupVar = group,
   repVar = rep,
   n_basis = 8,
   covType = "Homog_Uniform",
   corType = "exponential",
   returnFitted = TRUE,
   use_parallel = TRUE
 )

plot(fit)