R help - Aug 2006 - Help on function adf.test

Hello everybody,
   
  I've got a matrix called EUROPEDATA and I want to calculate the adf test
statistic (part of the tseries package) on a rolling basis for window my.win on
each column; i.e. each column of EUROPEDATA represents a particular variable;
for the first column I calculate the adf test statistic for window my.win = 60
for example, roll forward one observation, calculate the adf again, and so on,
until the end of the first column is reached and then I jump to the second
column etc. The code for doing this is given below:
   
  adfroll <- sapply(1:(ncol(EUROPEDATA)), function(i, my.data, my.win)
  {
  sss <- sapply(1:(nrow(my.data)-my.win), function (j, my.data, my.win)
  { my.data <- as.matrix(my.data)
  ans <- adf.test(na.omit(my.data[j:(j+my.win)]))
  return(ans$p.value)
  },my.data=my.data[,i],my.win=my.win) 
  },my.data=EUROPEDATA,my.win=60,simplify=T)
   
  The problem is that the adf test does not calculate this way. There is an
error saying:
   
  "Error in embed(y, k) : wrong embedding dimension"
   
  This error is generated from within the adf.test function. The embed function
is part of the stats package, which I load before doing the adf.test.
   
  I would be very obliged if anybody were to explain to me why this happens and
how I can correct it/ estimate what I want in a different way that will not
invoke this error.
   
  P.S. the function works fine if the adf.test on each column of the dataset is
calculated, that is, without the rolling window for each column:
   
   
  summaryadf <- sapply(1:ncol(EUROPEDATA),function(i,my.data)
  {
  tt <- adf.test(na.omit(my.data[,i]))
  return(tt$p.value)
  },my.data=EUROPEDATA) 
   
  Please note that the same error is generated with the pp.test function of the
tseries package
   
  Thanks in advance,
  Spyros

 		
---------------------------------

	[[alternative HTML version deleted]]

Put your time series into a ts or zoo object.  Now using
EuStockMarkets which is builtin data set in R.  (You might
want to use align = "right" in rollapply.)

library(tseries)
library(zoo)
eu91 <- window(EuStockMarkets, end = 1992)  # use portion for test data
eu91.p.value <- rollapply(eu91, 61, function(x) adf.test(x)$p.value)

On 8/28/06, Spiros Mesomeris <mesomeris at yahoo.co.uk>
wrote:
> Hello everybody,
>
>  I've got a matrix called EUROPEDATA and I want to calculate the adf
test statistic (part of the tseries package) on a rolling basis for window
my.win on each column; i.e. each column of EUROPEDATA represents a particular
variable; for the first column I calculate the adf test statistic for window
my.win = 60 for example, roll forward one observation, calculate the adf again,
and so on, until the end of the first column is reached and then I jump to the
second column etc. The code for doing this is given below:
>
>  adfroll <- sapply(1:(ncol(EUROPEDATA)), function(i, my.data, my.win)
>  {
>  sss <- sapply(1:(nrow(my.data)-my.win), function (j, my.data, my.win)
>  { my.data <- as.matrix(my.data)
>  ans <- adf.test(na.omit(my.data[j:(j+my.win)]))
>  return(ans$p.value)
>  },my.data=my.data[,i],my.win=my.win)
>  },my.data=EUROPEDATA,my.win=60,simplify=T)
>
>  The problem is that the adf test does not calculate this way. There is an
error saying:
>
>  "Error in embed(y, k) : wrong embedding dimension"
>
>  This error is generated from within the adf.test function. The embed
function is part of the stats package, which I load before doing the adf.test.
>
>  I would be very obliged if anybody were to explain to me why this happens
and how I can correct it/ estimate what I want in a different way that will not
invoke this error.
>
>  P.S. the function works fine if the adf.test on each column of the dataset
is calculated, that is, without the rolling window for each column:
>
>
>  summaryadf <- sapply(1:ncol(EUROPEDATA),function(i,my.data)
>  {
>  tt <- adf.test(na.omit(my.data[,i]))
>  return(tt$p.value)
>  },my.data=EUROPEDATA)
>
>  Please note that the same error is generated with the pp.test function of
the tseries package
>
>  Thanks in advance,
>  Spyros
>
>
> ---------------------------------
>
>        [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>