Dear R-helpers, I have 4-dimensional atmospheric data (x,y,z,t), which I want to analyse on spatio-temporal diversities. As far as I understand there only exists the possibility to construct time series as two-dimensional matrices (mts). For the moment, I hold it in different objects: 1. a four-dimensional array for the spatial related analyses 2. a two-dimensional mts timeserie, which was derived from 1. by computing spatial means. But, still this doesn't help for combined spatio-temporal analysis. One could regard the time dimension as just another linearily spaced dimension in the four-dimensional array, but when it comes analyses and graphics output it gets complicated, since one can't use all timeseries (ts-) related functions (require ts attributes). Could you provide me some comments, workarounds or your experiences on similar problems. Thanks in advance, Christian -------------- next part -------------- A non-text attachment was scrubbed... Name: christian.georges.vcf Type: text/x-vcard Size: 324 bytes Desc: Karte für Christian Georges Url : https://stat.ethz.ch/pipermail/r-help/attachments/20010906/bc572dbc/christian.georges.vcf
If your data is a bit like rainfall at multiple sites, I could probably offer some comments (assuming (x,y,z,t) is location and time?), although I don't know how useful they will be for your particular application. My (limited) experience is that R will do quite a lot up to bivariate time series (e.g. ccf function), but beyond that you need to write your own code. This could be in R, but may be more appropriate in a fully compiled language (like C or Fortran) if you have a large data set (because R is too slow for heavy number crunching work; although it appears to be faster than S-Plus). Again I'm not sure whether this is relevant for your data set, but a fairly common approach with climatological data is to attempt to reduce the dimensions using principle components. I understand PCA functions exist in R, although I've never used them. Depending on your data this approach may simplify the problem. -----Original Message----- From: Christian Georges [mailto:christian.georges at uibk.ac.at] Sent: Friday, 7 September 2001 02:48 To: r-help at stat.math.ethz.ch; Christian Georges Subject: [R] Array as time series? Dear R-helpers, I have 4-dimensional atmospheric data (x,y,z,t), which I want to analyse on spatio-temporal diversities. As far as I understand there only exists the possibility to construct time series as two-dimensional matrices (mts). For the moment, I hold it in different objects: 1. a four-dimensional array for the spatial related analyses 2. a two-dimensional mts timeserie, which was derived from 1. by computing spatial means. But, still this doesn't help for combined spatio-temporal analysis. One could regard the time dimension as just another linearily spaced dimension in the four-dimensional array, but when it comes analyses and graphics output it gets complicated, since one can't use all timeseries (ts-) related functions (require ts attributes). Could you provide me some comments, workarounds or your experiences on similar problems. Thanks in advance, Christian -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
A few indications about multivate analyses you can perform on time series:
- You can use any multivariate analysis like PCA and co, clustering methods
and so, considering time as just another variable. Although not implemented
in R, there are clustering methods with time connexity, i.e., in the final
dendrogram, all stations apppear in the chronological order and clusters are
constrained between adjacent stations in time (or space) only.
- A few multivariate analyses specific to space-time series:
+ Distogram
Mackas, D.L., 1984. Spatial autocorrelation of plankton community
composition in a continental shelf ecosystem. Limnol. Ecol., 20:451-471.
+ multiple autocorrelation with methods like crossD2 and D2 to the center.
Ibanez, F., 1981. Immediate detection of heterogeneities in continuous
multivariate oceanographic recordings. Application to time series analysis
of changes in the bay of Villefranche sur mer. Limnol. Oceanogr.,
26:336-349.
Ibanez, F., 1991. Treatment of the data deriving from the COST 647 project
on coastal benthic ecology: The within-site analysis. In: B. Keegan (ed),
Space and time series data analysis in coastal benthic ecology. Pp 5-43.
I certainly speak too much about something that is not finished yet... but
an R package for such multivariate time series analyses is in preparation
(due to the end of the year). See: http://www.sciviews.org/_passtec (in
French for the moment only).
All the best,
Philippe Grosjean
...........]<(({?<...............<?}))><...............................
) ) ) ) ) __ __
( ( ( ( ( |__) | _
) ) ) ) ) | hilippe |__)rosjean
( ( ( ( ( Marine Biol. Lab., ULB, Belgium
) ) ) ) ) __
( ( ( ( ( |\ /| |__)
) ) ) ) ) | \/ |ariculture & |__)iostatistics
( ( ( ( (
) ) ) ) ) e-mail: phgrosje at ulb.ac.be or phgrosjean at sciviews.org
( ( ( ( ( SciViews project coordinator (http://www.sciviews.org)
) ) ) ) ) tel: 00-32-2-650.29.70 (lab), 00-32-2-673.31.33 (home)
( ( ( ( (
) ) ) ) ) "I'm 100% confident that p is between 0 and 1"
( ( ( ( ( L. Gonick & W. Smith (1993)
) ) ) ) )
.......................................................................
>Dear R-helpers,
>I have 4-dimensional atmospheric data (x,y,z,t), which I want to analyse
>on spatio-temporal diversities.
>As far as I understand there only exists the possibility to construct
>time series as two-dimensional matrices (mts).
>For the moment, I hold it in different objects:
>1. a four-dimensional array for the spatial related analyses
>2. a two-dimensional mts timeserie, which was derived from 1. by
>computing spatial means.
>But, still this doesn't help for combined spatio-temporal analysis.
>One could regard the time dimension as just another linearily spaced
>dimension in the four-dimensional array, but when it comes analyses and
>graphics output it gets complicated, since one can't use all timeseries
>(ts-) related functions (require ts attributes).
>Could you provide me some comments, workarounds or your experiences on
>similar problems.
>Thanks in advance,
>Christian
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._