R/polyDlm.main.R

In dLagM: Time Series Regression Models with Distributed Lag Models

#' @export
polyDlm.main <- function(x , y , q , k , show.beta = TRUE ){
  n <- length(x)
  design  <- dlm(x = x , y = y , q = q )$designMatrix
  tr.matrix <- array(0, dim = c((q+1),(k+1)))
  design.z <- array(0,dim=c(length((q+1):n),(k+1)))
  design.z.colnames <- array(NA, (k+1))
  
  for (i in 1:(q+1)){
    s <- i - 1
    tr.matrix[i , 1] <- 1
    for (j in 2:(k+1)){
      tr.matrix[i , j] <- s^(j - 1)
    }
  }
  tr.matrix <- t(tr.matrix)
  
  for ( i in 1:nrow(design)){
    design.z[i,] <- t(tr.matrix %*% t(design[i,2:ncol(design)]))
  }
  
  for (i in 1:(k+1)){
    design.z.colnames[i] <- paste0("z.t",(i-1))
  }
  
  y.t <- design[,1]
  z <- data.frame(y.t, design.z)
  colnames(z) <- c("y.t" , design.z.colnames)
  model <- lm(y.t ~ ., data = z)
  model$call <- "Y ~ (Intercept) + X.t"
  if (show.beta == TRUE){
    tests <- poly.dlm.tests(coef = model$coefficients , tr.matrix = tr.matrix , y.t = y.t , design.z = design.z , n = n , q = q , print = TRUE )
    return(list(model = model, designMatrix = design.z, designMatrix.x = design, beta.coefficients = tests, transformation.matrix = tr.matrix))
  } else {
    return(list(model = model, designMatrix = design.z, designMatrix.x = design, transformation.matrix = tr.matrix))
  }
}