R/mnma.model.R

In vandy10s/mnma: Multivariate Network Meta-Analysis (NMA) within a Bayesian Framework

#' @title Create a data object for multivariate and univariate network meta-analysis (NMA)
#' @description The function generates a data object for \code{mnma.run} or \code{unma.run} operation, using outcome 1, outcome 2, and treatments inputs
#' @param outcome1 A data frame for outcome 1. First 2 (or 3) columns contain the information about number of events for each arm and the rest 2 (or 3) columns contain the sample size for each arm. Number of columns ranges from 2 to 3, depending on the number of maximum arms in the trial data.
#' @param outcome2 A data frame for outcome 2, which has the same dimensions as outcome1.
#' @param TX A data frame for tratments, which has maximum 3 treatment arms in 3 columns. The number of rows should coincide with outcome1 and outcome2. Each element needs to be numeric.
#' @param noa Maximum number of arms in the trial studies, which is either \code{2arm} or \code{3arm}. Note that number of arm is different from number of total treatments in the network.
#' @param ref Reference treatment in the analysis in a numeric class.
#' @param parameters List of parameters to infer in the analysis. They are either \code{simple} or \code{detail} depending on comparison type.
#' @export
#' @examples
#' data(acute)
#' # Split the dataset to create a mnma data object
#' outcome1 <- acute[,c(1:6)]     # outcome1
#' outcome2 <- acute[,c(7:12)]    # outcome2
#' TX    <- acute[,13:15]
#' # Convert them into log odds ratio inputs
#' data <- mnma.model(outcome1,outcome2,TX, ref=1, parameters="simple")
#' data
#'
#' @details A \code{simple} list for parameters includes log odds ratio (comparison) parameters between reference and all other treatments. A \code{detail} list for parameters includes log odds ratio parameters between all treatments. Both cases include a correlation parameter and heterogeneity parameters for two outcomes.
#' @return An object of class \code{mnma.model} that contains data information and parameter lists to estimate for multivariate and univariate network meta-analysis.


mnma.model <- function(outcome1, outcome2, TX ,noa="3arm", ref=1, parameters="simple"){

  mulv <- mnma:::mul.model(data1=outcome1, data2=outcome2, TX=TX, ref=ref, noa=noa, parameters=parameters)
  univ <- mnma:::uni.model(data1=outcome1, data2=outcome2, TX=TX, ref=ref, noa=noa, parameters=parameters)

  source <- list(multivariate=mulv, univariate=univ)
  class(source) <- "mnma.model"
  return(source)

}