CODEWITHSUNDEEP
rtime-series
Rajan
Rajan

Reputation: 463

arima method in mtsdi

I have a large data set(more than 2000 rows and 2000 variables) with lots of missing values. I am using mnimputfunction of mtsdi package of R for imputing all missing values. This is my code

formula = data
imput_out <- mnimput(formula,data, by = NULL, log = FALSE, log.offset = 1,
                 eps = 1e-3, maxit = 1e2, ts = TRUE, method = "arima", ar.control = list(order = c(1,1,1), period = 4,  f.eps = 1e-6, f.maxit = 1e3, ga.bf.eps = 1e-6,verbose = TRUE, digits = getOption("digits")))

But I am getting an error 

Error in o[1:3, j] : incorrect number of dimensions

Please help me out.

Upvotes: 4

Views: 1404

Answers (1)

hoshie
hoshie

Reputation: 76

you have to get real deep into the package source to uncover whats going on here. the ar.control is placed into a variable o that is iterated on by the j # of columns that you put into your formula. so if your formula looks like ~c31+c32+c33 your ar term need to be 3 columns of (p,d,q) values

I assigned it outside of the ar.control parameter for ease of editing arcontrol<-list(order=cbind(c(1,0,0),c(0,0,1),c(1,0,0)), period=NULL) 

mnimput(formula,data,eps=1e-3,ts=TRUE, method="arima", ar.control=arcontrol

here is the package source if you are interested

function (xn, o, s, eps, maxit) 
{
  rows <- dim(xn)[1]
  cols <- dim(xn)[2]
  models <- as.list(rep(NA, cols))
  ar.pred <- matrix(NA, nrow = rows, ncol = cols)
  for (j in 1:cols) {
    if (is.null(s)) {
      order <- o[1:3, j]
      seasonal <- list(order = c(0, 0, 0), period = NA)
    }
    else {
      order <- o[1:3, j]
      seasonal <- list(order = o[4:6, j], period = s)
    }
    models[[j]] <- arima(xn[, j], order = order, seasonal = seasonal, 
      xreg = NULL, optim.control = list(maxit = maxit, 
        reltol = eps))
    ar.pred[, j] <- xn[, j] - residuals(models[[j]])
  }
  retval <- list(ar.pred = ar.pred, models = models)
  return(retval)
}

Upvotes: 6

Related Questions