R/readZ.R

In CorReg: Linear Regression Based on Linear Structure Between Variables

#' Read the structure and explain it
#' 
#' @description This function describes the structure of sub-regression given by an adjacency matrix. It computes the associated regression coefficients and R-squared for each sub-regression.
#' 
#' @param Z binary adjacency matrix of the structure (size p)
#' @param B is the complete structure (Z with sub-regression coefficients instead of 1 and an additional first line for the intercepts)
#' @param crit define the criterion to use: c("none","R2","F","sigmaX")
#' @param varnames the names of the variables (size p)
#' @param output indicates the content of the output: c("index","names","all")
#' @param X is a data frame or matrix containing the dataset
#' @param order define the order used (0: none, -1: decreasing, 1: growing) for printing
#' 
#' @return a list containing the sub-regressions details. Each item of the list represents
#' a subregression. First element is the R-square.Second element is the variable that is regressed by others.
#' Then comes the list of the explicative variables in the subgression and the associated coefficients (in the first column).
#' 
#' @examples
#' data <- mtcars
#' # we first search a sub-regression structure
#' res <- structureFinder(X = data, nbini = 30, verbose = 0)
#' # then we can try to interpret it
#' readZ(Z = res$Z_opt, crit = "R2", output = "all", X = data)
#' # each component is a sub-regression
#' # First line: The adjusted R-squared is given
#' # Second line: the name of the covariate that is regressed by others
#' # other lines: Coefficients of sub-regression and name of the associated covariate
#' 
#' @export
readZ <- function(Z = Z, B = NULL, crit = c("none", "R2", "F", "sigmaX"), varnames = NULL, output = c("index", "names", "all"), X = NULL, order = 1) {
  p = ncol(Z)
  output = match.arg(output)
  crit = match.arg(crit)
  res = list()
  if (output != "index" & is.null(varnames)) { # if names needed but unknown
    if (length(names(X)) != p) {
      varnames = 1:p
    } else {
      varnames = names(X)
    }
  }
  if (is.null(B) & (crit != "none" | output == "all")) { # if B needed but unknown
     B = hatB(Z = Z, X = X)
  }
  if (crit != "none") {
    R2vect = R2Z(Z = Z, X = X, adj = TRUE, crit = crit)
    R2vect = R2vect[R2vect != 0]
    quiI2 = which(colSums(Z) != 0)
    if (order == 1) { # increasing order
       neworder = order(R2vect, decreasing = FALSE)

    } else if (order == -1) {
       neworder = order(R2vect, decreasing = TRUE)

    } else {
       neworder = 1:length(quiI2)
    }
    if (output == "index") {
      for (i in 1:length(quiI2)) {
        res[[i]] = c(quiI2[neworder[i]], which(Z[, quiI2[neworder[i]]] != 0))
      }
    } else if (output == "names") {
      for (i in 1:length(quiI2)) {
        res[[i]] = varnames[c(quiI2[neworder[i]], which(Z[, quiI2[neworder[i]]] != 0))]
      }
    } else { # all
      for (i in 1:length(quiI2)) {
        beta = B[, quiI2[neworder[i]]]
        ssreg = data.frame(cbind(c(R2vect[neworder[i]], NA, beta[beta != 0]), c(crit, varnames[quiI2[neworder[i]]], "intercept", varnames[which(Z[, quiI2[neworder[i]]] != 0)])))
        names(ssreg) = c("coefs", "var")
        res[[i]] = ssreg
      }
    }
  } else {
    quiI2 = which(colSums(Z) != 0)
    if (output == "index") {
      for (i in 1:length(quiI2)) {
        res[[i]] = c(quiI2[i], which(Z[, quiI2[i]] != 0))
      }
    } else if (output == "names") {
      for (i in 1:length(quiI2)) {
        res[[i]] = varnames[c(quiI2[i], which(Z[, quiI2[i]] != 0))]
      }
    } else { # all
      for (i in 1:length(quiI2)) {
        beta = B[, quiI2[i]]
        ssreg = data.frame(cbind(c(NA, beta[beta != 0]), c(varnames[quiI2[i]], "intercept", varnames[which(Z[, quiI2[i]] != 0)])))
        names(ssreg) = c("coefs", "var")
        res[[i]] = ssreg
      }
    }
  }
  return(res)
}