[R] Help with Sequential Differencing

[Sarah Goslee]

I missed something in my original reply:

You intended a recursive approach, I think, but that wasn't clear in
your code, and I simply repeated your actual code rather than your
intended code in my response. If that's what you want, you need
something more like:

DCred<- list()
DCred[[1]] <- diff(Cred, difference=1)

for(i in 2:5){
  print(DCred[[i]]<- diff(DCred[[i-1]], lag=i, difference=1))
}

Doing it with separately-named files will be much more complicated,
something like

DCred1 <- diff(Cred, difference=1)

for(i in 2:5){
  print(assign(paste("DCred", i, sep=""), diff(get(paste("DCred", i-1,
sep="")), lag=i, difference=1))
}

(both are untested)

You might also benefit from reading Intro to R, which covers list structures.

Sarah

On Fri, Jun 15, 2012 at 9:22 AM, Lekgatlhamang, lexi Setlhare
<lexisetlhare at yahoo.com> wrote:
> Hi Sarah,
>
> Sorry for taking more of your time. I thought that before investigating the
> use of assign and paste, I should first understand the approach of using
> 'list()'. However, it seems to me that the use of list does not work very
> well, unless I have wrongly implemented your suggestion. I implemented your
> suggestion as follows:
> DCred<- list()
> for(i in 1:5){
> print(DCred[[i]]<- diff(DCred, lag=i, difference=1))
> }
> Upon execution of the loop, the following was printed out:
> list()
> list()
> list()
> list()
> list()
> After the loop I tried to access the values for DCred1, DCred2, etc but to
> no avail. See my commands and the associated output below:
>
>> DCred[[1]]
> list()
>> DCred[[i]]
> list()
>> DCred[1]
> [[1]]
> list()
> Anyway, I then calculate DCred in order to see if the regression
>
> DCred <-lm(DCred~DCred1+DCred2)
>
> will work.
> But it did not. Instead I got an error message - see below.
>
>> model<- lm(DCred~DCred[1]+DCred[2], drop.unused.levels = TRUE)
> Error in model.frame.default(formula = DCred ~ DCred[1] + DCred[2],
> drop.unused.levels = TRUE) :
>   variable lengths differ (found for 'DCred[1]')
>
> I then modified the right-hand-side of the equation, but still it did not
> work:
>
>> model <- lm(DCred~DCred[[1]])
> Error in model.frame.default(formula = DCred ~ DCred[[1]],
> drop.unused.levels = TRUE) :
>   variable lengths differ (found for 'DCred[[1]]')
>
> I will remain reading up on the assign and paste commands.
> Thanks. Lexi
> From: Sarah Goslee <sarah.goslee at gmail.com>
> To: "Lekgatlhamang, lexi Setlhare" <lexisetlhare at yahoo.com>
> Cc: "r-help at r-project.org" <r-help at r-project.org>
> Sent: Friday, June 15, 2012 12:54 PM
> Subject: Re: [R] Help with Sequential Differencing
>
> If you really want DCred1, DCred2, etc, you'll need to use assign() and
> paste() within your loop. But he's a more R-ish way:
>
> # Differences of Cred
>
> DCred <- list()
> for(i in 1:5){
> print(DCred[[i]]<- diff(DCred, lag=i, difference=1))
>
> }
>
> DCred[[1]]
>
> Sarah
>
> On Friday, June 15, 2012, Lekgatlhamang, lexi Setlhare wrote:
>
> Dear R Users,
> Sorry for what seems like I am re-posting. When I was typing my initial
> posting, I intended to copy and paste the commands from my script, but ended
> up forgetting. I am now pasting the commands in this email.NB: Below is a
> copy of 'all' the relevant commands in my script
>
>  ##################################################
> data<- read.table("C:\\myfiles\\from drive D\\myfiles 080208\\An R Test
> folder\\Dataset2a.csv", head=TRUE, sep=",")
> str(data)
> data
>
> # Extract the variables in data, rename and difference them
>
> Cred <- ts(data[, 2], frequency=12, start=c(2001, 1)) # Cred denotes Credit
>
> # Differences of Cred
> DCred<- diff(Cred, difference=1)
> DCred
> for(i in 1:5){
> print(DCred[i]<- diff(DCred, lag=i, difference=1))
>
> }
>
> DCred[1]
>
>  DCred[2]
>
>  DCred[3]
>
>  ################################################
> Thanks Lexi
>
>
> ________________________________
>
> To: "r-help at r-project.org" <r-help at r-project.org>
> Sent: Friday, June 15, 2012 11:39 AM
> Subject: Help with Sequential Differencing
>
>
>
>
> Dear R Users,
>
> I have struggled with the following problem for days, which I thought was
> simple, although it would likely be basic to most of you.
> I am working with time series data.
> In my script, my intention is to create first differences of the variables
> in the file so that I end up estimating an equation of the form:
> DCred(t) =c +
> DCred(t-1)+DCred(t-2)+...+DBoB(t)+DBoB(t-1)+DBoB(t-2)+...+Drvr(t)+Drvr(t-1)+Drvr(t-2)+...+e(t)
>
> Where D at the beginning of each variable represents 'change', for first
> difference and e(t) is the error term.
> Now I am trying to use loops to calculate 5 lagged first-differences of each
> variable in the dataset - e.g., DCred(t-1), DCred(t-2), ..., DCred(t-5).
> Example:
> # Differences of Cred
> DCred<- diff(Cred, difference=1)
> DCred
> for(i in 1:5){
> print(DCred[i]<- diff(DCred, lag=i, difference=1))
> }
> After I calculated the contemporaneous first difference DCred, this loop is
> meant to compute the subsequent first differences of the same variable;
> i.e., DCred(t-1) and call it DCred1, DCred(t-2) and call it DCred2, ... and
> DCred(t-5) and call it DCred5.
> The loop works, at least I think so. But now after the loop has executed,
> when I type DCred1[1] (which I thought would give me the first value in the
> series for DCred(t-1)), called DCred1, I get a message "object 'DCred1' not
> found". Similarly typing Dcred1[2] (which I thought would give the second
> value of DCred(t-1)), ie., the second value of DCred1, gives "object
> DCred1[2] not found", etc.
> A copy of the commands and error messages is below:
>> DCred1[1]
> Error: object 'DCred1' not found
>> DCred1[2]
> Error: object 'DCred1' not found
>
> How can I solve this problem? Thank you kindly for your time.



-- 
Sarah Goslee
http://www.functionaldiversity.org