R/banerjee.R

In combcoint: A Joint Test-Statistic for the Null of Non-Cointegration

#' @describeIn englegranger Executes Banerjee Test.
#' @export
banerjee <- function(formula, data, lags = 1, trend = "const"){

  # ---- Check Syntax ----
  mf <- match.call()
  m <- match(c("formula", "data"), names(mf), 0L)
  if (is.null(data))
    stop()
  mf <- mf[c(1L, m)]
  mf[[1L]] <- quote(stats::model.frame)
  mf <- eval(mf, parent.frame())
  mt <- attr(mf, "terms")
  mf <- stats::na.omit(mf)
  y <- stats::model.response(mf, "numeric")
  x <- stats::model.matrix(mt, mf)[, -1]
  trend <- match.arg(trend,
                     choices = c("none", "const", "trend"))

  lag <- lags
  if (lag < 0)
    stop("Lags must be set to a non negative value.")


  # ---- Lag Matrix ----
  Xlag <- cbind(y, x)
  Y_dif <- diff(y) # muss als numeric vorliegen
  W <- diff(x)
  Xlag_diff <- as.data.frame(diff(Xlag))

  lag_mat <- matrix(NA, nrow = nrow(Xlag_diff), ncol = lags)
  lag_vec <- c()

  if (lags >= 1) {
    for (j in 1:ncol(Xlag_diff)) {
      for (i in 1:lags) {
        lag_mat[, i] <- Hmisc::Lag(Xlag_diff[, j], shift = i)
      }
      lag_vec <- cbind(lag_vec, lag_mat)
    }
    W <- cbind(W, lag_vec)
  }


  # ---- Banerjee Test ----
  res <- matrix(NA, nrow = nrow(Xlag) - lags - 1, ncol = ncol(Xlag))

  if (identical(trend, "none")) {
    for (i in 1:ncol(Xlag)) {
      loop_lm <- stats::lm(Hmisc::Lag(Xlag[, i], shift = 1)[-1] ~ W - 1)
      res[, i] <- as.numeric(loop_lm$residuals)
    }
  } else if (identical(trend, "const")) {
    for (i in 1:ncol(Xlag)) {
      loop_lm <- stats::lm(Hmisc::Lag(Xlag[, i], shift = 1)[-1] ~ W)
      res[, i] <- as.numeric(loop_lm$residuals)
    }
  } else if (identical(trend, "trend")) {
    tr <- seq_along(y)[-1]
    W <- cbind(W, tr)
    for (i in 1:ncol(Xlag)) {
      loop_lm <- stats::lm(Hmisc::Lag(Xlag[, i], shift = 1)[-1] ~ W)
      res[, i] <- as.numeric(loop_lm$residuals)
    }
    W <- W[, -which(colnames(W) == "tr")]
  }

  if (identical(trend, "none")) {
    BB_lm <- stats::lm(Y_dif ~ W - 1)
  } else if (identical(trend, "const")) {
    BB_lm <- stats::lm(Y_dif ~ W)
  } else if (identical(trend, "trend")) {
    tr <- seq_along(y)[-1]
    W <- cbind(W, tr)
    BB_lm <- stats::lm(Y_dif ~ W)
    W <- W[, -which(colnames(W) == "tr")]
  }

  BB_res <- BB_lm$residuals

  if (identical(trend, "none")) {
    lm_res <- stats::lm(BB_res ~ res -1)
  } else if (identical(trend, "const")) {
    lm_res <- stats::lm(BB_res ~ res)
  } else if (identical(trend, "trend")) {
    tr <- seq_along(y)[-c(1, 2)]
    suppressWarnings(res <- cbind(res, tr))
    lm_res <- stats::lm(BB_res ~ res)
    res <- res[, -which(colnames(res) == "tr")]
  }

  betas <- stats::coef(lm_res)
  var_mat <- stats::vcov(lm_res)
  if (identical(trend, "none")) {
    test.stat <- as.numeric(betas[1]/sqrt(var_mat[1, 1]))
  } else {
  test.stat <- as.numeric(betas[2]/sqrt(var_mat[2, 2]))
  }

  out <- list(test.stat = test.stat,
              lags = lags,
              trend = trend,
              betas = betas,
              var.cov = var_mat,
              test = "Banerjee",
              formula = formula)
  return(out)
}