[R] Cross Spectrum Analysis

[Maura E Monville]

R function "spectrum" expects a time series as input. I have attached a
compressed archive with two detrended and denoised signals (txt format)
whose spectra I would like to compare.
I start out trying to generate a multivariate time series.
Please, notice the different signals length. Moreover, R command "cbind"
forces a matrix with number of rows equal to the longer of the two signals,
padding the shorter one by replicating it from its start values.... What
happens in the frequency domain ?
The signals are sampled at 30 Hz.

> s10146 <- read.table("10146-Clean-Signal.txt")
> s45533 <- read.table("45533-Clean-Signal.txt")
> v10146 <- as.vector(s10146[,1])
> length(v10146)
[1] 8133
> v45533 <- as.vector(s45533[,1])
> length(v45533)
[1] 6764
> xx <-cbind(v10146, v45533)
> dim(xx)
[1] 8133    2
> v45533[1:10]
 [1] -1.7721546 -1.7482835 -1.6964711 -1.6154405 -1.5045701 -1.3747449
 [7] -1.2332980 -1.0912172 -0.9585821 -0.8420886
> xx[6760:6770,]
          v10146     v45533
 [1,] -0.8585375 -0.6076069
 [2,] -0.8060065 -0.5288312
 [3,] -0.7541174 -0.4447711
 [4,] -0.7028816 -0.3592778
 [5,] -0.6524279 -0.2767786
 [6,] -0.6027233 -1.7721546    # start replicating shorter signal
 [7,] -0.5536868 -1.7482835
 [8,] -0.5052780 -1.6964711
 [9,] -0.4574095 -1.6154405
[10,] -0.4097922 -1.5045701
[11,] -0.3623641 -1.3747449

> twosig <- ts(xx,deltat=0.033,start=0)  # time series







On Sun, May 4, 2008 at 2:14 PM, stephen sefick <ssefick at gmail.com> wrote:

> is this a problem?  are there error messages?  if so could you provide
> them.  Try
> as.matrix(yourdata).  One thing you could do is create a moving average
> that reduces the signals to the lowest common denominator.  Could you
> provide reproducable code with maybe a toy data set so anybody could have a
> look at what is going on.
> good luck
>
> Stephen
>
>
>