[R] Embed function strips out date index

[Peter Ehlers]

On 2010-06-30 14:32, Manussawee Sukunta wrote:
> Hi,
>
> I'm having especially hard time today and couldn't find any
> clue/answer through the internet.  I hope you can help.
>
> I'm in a process of writing a script to estimate error correction
> model, and I was following an example in Bernhard Pfaff's Analysis of
> Integrated and Cointegrated Time Series with R.  I have the following
> price data:
>
>> head(series,15)
>             PX_SETTLE PX_SETTLE.1
> 2009-01-02    4515.0      925.50
> 2009-01-05    4540.5      927.50
> 2009-01-06    4603.5      930.50
> 2009-01-07    4470.5      905.25
> 2009-01-08    4474.5      906.75
> 2009-01-09    4430.5      885.50
> 2009-01-12    4402.0      868.00
> 2009-01-13    4343.5      868.50
> 2009-01-14    4130.5      839.75
> 2009-01-15    4070.5      839.25
> 2009-01-16    4129.5      848.50
> 2009-01-20    4032.0      806.00
> 2009-01-21    4018.0      836.75
> 2009-01-22    4011.0      825.50
> 2009-01-23    3998.0      823.50
>
> Then I defined
> series.d = embed(diff(series),dim=2)
>
> which resulted in
>> head(series.d,15)
>          [,1]   [,2]   [,3]   [,4]
>   [1,]   25.5   2.00     NA     NA
>   [2,]   63.0   3.00   25.5   2.00
>   [3,] -133.0 -25.25   63.0   3.00
>   [4,]    4.0   1.50 -133.0 -25.25
>   [5,]  -44.0 -21.25    4.0   1.50
>   [6,]  -28.5 -17.50  -44.0 -21.25
>   [7,]  -58.5   0.50  -28.5 -17.50
>   [8,] -213.0 -28.75  -58.5   0.50
>   [9,]  -60.0  -0.50 -213.0 -28.75
> [10,]   59.0   9.25  -60.0  -0.50
> [11,]  -97.5 -42.50   59.0   9.25
> [12,]  -14.0  30.75  -97.5 -42.50
> [13,]   -7.0 -11.25  -14.0  30.75
> [14,]  -13.0  -2.00   -7.0 -11.25
> [15,]  169.0   7.25  -13.0  -2.00
>
> The new data series.d now has no date index.  I'm not sure how to get
> it back.  I tried to xts -->  order.by = index(series), but the vector
> lengths are now not the same.  I feel like the answer might be
> obvious, but I just can't see it.  Again, I tried searching various
> forums and sites, but I couldn't find my answer.  I feel like I'm just
> going around a circle.  I hope someone can help me and shed some light
> on this problem.

It looks like series is a dataframe with your dates
as rownames. If you want series.d (a matrix) to have
the same rownames:

  rownames(series.d) <- rownames(series)

But I'm not sure how much sense that makes, especially
for columns 3 and 4.

   -Peter Ehlers

>
> Thank you,
> Manussawee
>