[R] Determine the dimension-names of an element in an array in R

Poersching Poersching79 at web.de
Sat Aug 1 19:00:12 CEST 2009


Hey,
oh yes, but now I have realy the ultimate solution... ;-)
Here it comes:

a= c("A1","A2","A3","A4","A5")
b= c("B1","B2","B3")
c= c("C1","C2","C3","C4")
d= c("D1","D2")
e= c("E1","E2","E3","E4","E5","E6","E7","E8")

DataArray_1 = array(c(rnorm(240)),dim=c(length(a),length(b),
 length(d),length(e)),dimnames=list(a,b,d,e))
DataArray_2 = array(c(rnorm(320)), dim=c(length(a),length(c),
 length(d),length(e)),dimnames=list(a,c,d,e))

z<-apply(as.matrix(a),c(1,2),function(f1)
    apply(as.matrix(d),c(1,2),function(f2)
   
apply(DataArray_1[dimnames(DataArray_1)[[1]]==f1,,dimnames(DataArray_1)[[3]]==f2,],1,
function(d1)
   
apply(DataArray_2[dimnames(DataArray_2)[[1]]==f1,,dimnames(DataArray_2)[[3]]==f2,],1,
function(d2)
    cor(d1,d2))
)))
Correl = array(z, dim=c(length(c),length(b),
 length(d),length(a)),dimnames=list(c,b,d,a))
Correl<-aperm(Correl,c(4,2,1,3))

So, best Regards,
Christian

Sauvik De schrieb:
> Hi Christian:
>
> Many thank for the code.
>
> But I am afraid that your code still has a problem in terms of
> providing correct correlation. For example, if you look at the
> correlation between DataArray_1["A2","B1","D1",] and
> DataArray_2["A2","C1","D1",] after running your code, you will notice
> that this is actually the correlation between
> DataArray_1["A2","B1","D1",] and DataArray_2["A1",C1","D1",] and so on.
>
> The code gives the correct result only in case where elements
> corresponding to "A1" & "D1" are involved in DataArray_1 & DataArray_2.
>
> The problem is in
>
> Correl<-Correl[1:length(c),,,]
>
> We need to select elements of Correl more carefully to reach a proper
> solution.
>
> Thanks,
> Sauvik
>
>
> On Wed, Jul 29, 2009 at 11:41 PM, Poersching <Poersching79 at web.de
> <mailto:Poersching79 at web.de>> wrote:
>
>     Hey,
>     i have forgotten to generalize the code so
>
>     Correl<-Correl[1:4,,,]
>
>     must be
>
>     Correl<-Correl[1:length(c),,,]
>
>     it's because the comparison levels. I think you don't want the
>     correlation betweeen A1, B1, D1 and A2, C1, D1 ,
>     but between A1, B1, D1 and A1, C1, D1 or between A1, B1, D1 and
>     A1, C2, D1.
>     So the "1:length(c)" writes only the correlation between the B and
>     C out
>     of the whole correlation array.
>     That's also why the sequence in the second apply function is changed.
>
>     Regards Christian.
>
>     Poersching schrieb:
>     > Hey,
>     > I think I have a solution for your problem:
>     >
>     > Correl<-apply(DataArray_1,1:3, function(d1)
>     >   apply(DataArray_2,c(2,1,3), function(d) cor(d1,d))
>     > )
>     > Correl<-Correl[1:4,,,]
>     > dimnames(Correl)[[1]]<-c
>     > Correl<-aperm(Correl,c(2,3,1,4))
>     >
>     > This one should work. :-)
>     >
>     > Best Regards,
>     > Christian
>     >
>     > Sauvik De schrieb:
>     >
>     >> Hi there,
>     >>
>     >> Thanks again for your reply. I know for-loop is always a
>     solution to
>     >> my problem and I had already coded using for-loop. But the
>     number of
>     >> levels for each dimension is large enough in actual problem and
>     hence
>     >> it was time-consuming.
>     >> So, I was just wondering if there are any other alternative
>     way-outs
>     >> to solving my problem. That's why I tried with apply functions
>     >> (sapply)assuming that this might work out faster even
>     fractionally as
>     >> compared to for-loop.
>     >>
>     >> Cheers,
>     >> Sauvik
>     >>
>     >> On Mon, Jul 27, 2009 at 12:28 AM, Poersching
>     <Poersching79 at web.de <mailto:Poersching79 at web.de>
>     >> <mailto:Poersching79 at web.de <mailto:Poersching79 at web.de>>> wrote:
>     >>
>     >>     Sauvik De schrieb:
>     >>
>     >>>     Hi:
>     >>>     Lots of thanks for your valuable time!
>     >>>
>     >>>     But I am not sure how you would like to use the function
>     in this
>     >>>     situation.
>     >>>
>     >>>     As I had mentioned that the first element of my output array
>     >>>     should be like:
>     >>>
>     >>>    
>     cor(DataArray_1[dimnames(Correl)[[1]][1],dimnames(Correl)[[2]][1],dimnames(Correl)[[4]][1],],DataArray_2[dimnames(Correl)[[1]][1],dimnames(Correl)[[3]][1],dimnames(Correl)[[4]][1],],use="pairwise.complete.obs")
>     >>>
>     >>>     in my below code.
>     >>>
>     >>>     and
>     >>>
>     >>>     the output array of correlation I wish to get using
>     "sapply" as
>     >>>     follows:
>     >>>
>     >>>     Correl = sapply(Correl,function(d)
>     >>>     cor(DataArray_1[...],DataArray_2[...],
>     >>>     use="pairwise.complete.obs"))
>     >>>
>     >>>     So it would be of great help if you could kindly specify
>     how to
>     >>>     utilise your function "findIndex" in ...
>     >>>
>     >>>     Apologies for all this!
>     >>>
>     >>>     Thanks & Regards,
>     >>>     Sauvik
>     >>>
>     >>>
>     >>     Hey,
>     >>     sorry, I haven't understood your problem last time, but now
>     this
>     >>     solution should solve your problem, so I hope. :-)
>     >>     It's only a for to loop, but an apply function may work too. I
>     >>     will think about this, but for now...  ;-)
>     >>
>     >>     la<-length(a)
>     >>     lb<-length(b)
>     >>     lc<-length(c)
>     >>     ld<-length(d)
>     >>     for (ia in 1:la) {
>     >>       for (ib in 1:lb) {
>     >>         for (ic in 1:lc) {
>     >>           for (id in 1:ld) {
>     >>             Correl[ia,ib,ic,id]<-cor(
>     >>              DataArray_1[dimnames(Correl)[[1]][ia],
>     >>              dimnames(Correl)[[2]][ib],
>     >>              dimnames(Correl)[[4]][id],]
>     >>              ,
>     >>              DataArray_2[dimnames(Correl)[[1]][ia],
>     >>               dimnames(Correl)[[3]][ic],
>     >>               dimnames(Correl)[[4]][id],]
>     >>              ,
>     >>              use="pairwise.complete.obs")
>     >>           }
>     >>         }
>     >>       }
>     >>     }
>     >>     ## with function findIndex you can find the dimensions with
>     >>     ## i.e. cor values greater 0.5 or smaller -0.5, like:
>     >>     findIndex(Correl,Correl[Correl>0.5])
>     >>     findIndex(Correl,Correl[Correl<(-0.5)])
>     >>
>     >>     I have changed the code of the function findIndex in line which
>     >>     contents: el[j]<-which(is.element(data,element[j]))
>     >>
>     >>     Rigards,
>     >>     Christian
>     >>
>     >>
>     >>>     On Sun, Jul 26, 2009 at 3:54 PM,
>     Poersching<Poersching79 at web.de <mailto:Poersching79 at web.de>
>     >>>     <mailto:Poersching79 at web.de <mailto:Poersching79 at web.de>>>
>     wrote:
>     >>>     > Sauvik De schrieb:
>     >>>     >
>     >>>     > Hi Gabor:
>     >>>     > Many thanks for your prompt reply!
>     >>>     > The code is fine. But I need it in more general form as
>     I had
>     >>>     mentioned that
>     >>>     > I need to input any 0 to find its dimension-names.
>     >>>     >
>     >>>     > Actually, I was using "sapply" to calculate correlation and
>     >>>     this idea was
>     >>>     > required in the middle of correlation calculation.
>     >>>     > I am providing the way I tried my calculation.
>     >>>     >
>     >>>     > a= c("A1","A2","A3","A4","A5")
>     >>>     > b= c("B1","B2","B3")
>     >>>     > c= c("C1","C2","C3","C4")
>     >>>     > d= c("D1","D2")
>     >>>     > e= c("E1","E2","E3","E4","E5","E6","E7","E8")
>     >>>     >
>     >>>     > DataArray_1 = array(c(rnorm(240)),dim=c(length(a),length(b),
>     >>>     > length(d),length(e)),dimnames=list(a,b,d,e))
>     >>>     > DataArray_2 = array(c(rnorm(320)),
>     dim=c(length(a),length(c),
>     >>>     > length(d),length(e)),dimnames=list(a,c,d,e))
>     >>>     >
>     >>>     > #Defining an empty array which will contain the correlation
>     >>>     values (output
>     >>>     > array)
>     >>>     > Correl = array(NA, dim=c(length(a),length(b),
>     >>>     > length(c),length(d)),dimnames=list(a,b,c,d))
>     >>>     >
>     >>>     > #Calculating Correlation between attributes b & c over
>     values of e
>     >>>     > Correl = sapply(Correl,function(d)
>     >>>     cor(DataArray_1[...],DataArray_2[...],
>     >>>     > use="pairwise.complete.obs"))
>     >>>     >
>     >>>     > This is where I get stuck.
>     >>>     > In the above, d is acting as an element in the "Correl"
>     array.
>     >>>     Hence I need
>     >>>     > to get the dimension-names for d.
>     >>>     >
>     >>>     > #The first element of Correl will be:
>     >>>     >
>     >>>    
>     cor(DataArray_1[dimnames(Correl)[[1]][1],dimnames(Correl)[[2]][1],dimnames(Correl)[[4]][1],],DataArray_2[dimnames(Correl)[[1]][1],dimnames(Correl)[[3]][1],dimnames(Correl)[[4]][1],],use="pairwise.complete.obs")
>     >>>     >
>     >>>     > So my problem boils down to extracting the dim-names in
>     terms
>     >>>     of element(d)
>     >>>     > and not in terms of Correl (that I have mentioned as
>     "..." in
>     >>>     the above
>     >>>     > code)
>     >>>     >
>     >>>     > My sincere thanks for your valuable time & suggestions.
>     >>>     >
>     >>>     > Many Thanks & Kind Regards,
>     >>>     > Sauvik
>     >>>     >
>     >>>     >
>     >>>     > On Sun, Jul 26, 2009 at 5:26 AM, Gabor Grothendieck
>     >>>     <ggrothendieck at gmail.com <mailto:ggrothendieck at gmail.com>
>     <mailto:ggrothendieck at gmail.com <mailto:ggrothendieck at gmail.com>>
>     >>>     >
>     >>>     >
>     >>>     > wrote:
>     >>>     >
>     >>>     >
>     >>>     >
>     >>>     >
>     >>>     > Try this:
>     >>>     >
>     >>>     >
>     >>>     >
>     >>>     > ix <- c(1, 3, 4, 2)
>     >>>     > mapply("[", dimnames(mydatastructure), ix)
>     >>>     >
>     >>>     >
>     >>>     > [1] "S1" "T3" "U4" "V2"
>     >>>     >
>     >>>     >
>     >>>     > On Sat, Jul 25, 2009 at 5:12 PM, Sauvik
>     >>>     De<sauvik.stat at gmail.com <mailto:sauvik.stat at gmail.com>
>     <mailto:sauvik.stat at gmail.com <mailto:sauvik.stat at gmail.com>>> wrote:
>     >>>     >
>     >>>     >
>     >>>     > Hi:
>     >>>     > How can I extract the dimension-names of a pre-defined
>     element in a
>     >>>     > multidimensional array in R ?
>     >>>     >
>     >>>     > A toy example is provided below:
>     >>>     > I have a 4-dimensional array with each dimension having
>     certain
>     >>>     length.
>     >>>     >
>     >>>     >
>     >>>     > In
>     >>>     >
>     >>>     >
>     >>>     > the below example, "mydatastructure" explains the
>     structure of
>     >>>     my data.
>     >>>     >
>     >>>     > mydatastructure = array(0,
>     >>>     >
>     >>>     >
>     >>>     > dim=c(length(b),length(z),length(x),length(d)),
>     >>>     >
>     >>>     >
>     >>>     > dimnames=list(b,z,x,d))
>     >>>     >
>     >>>     > where,
>     >>>     > b=c("S1","S2","S3","S4","S5")
>     >>>     > z=c("T1","T2", "T3")
>     >>>     > x=c("U1","U2","U3","U4")
>     >>>     > d=c("V1","V2")
>     >>>     >
>     >>>     > Clearly, "mydatastructure" contains many 0's.
>     >>>     > Now how can I get the dimension-names of any particular 0 ?
>     >>>     > That is, my input should be a particular 0 in the array
>     >>>     "mydatastructure"
>     >>>     > (Suppose this 0 corresponds to S1,T3,U4 & V2 in the
>     array). Then my
>     >>>     >
>     >>>     >
>     >>>     > output
>     >>>     >
>     >>>     >
>     >>>     > should be S1,T3,U4 & V2.
>     >>>     >
>     >>>     > The function "dimnames" didn't help me with the solution.
>     >>>     > Any idea will greatly be appreciated.
>     >>>     >
>     >>>     > Thanks for your time!
>     >>>     >
>     >>>     > Kind Regards,
>     >>>     > Sauvik
>     >>>     >
>     >>>     >        [[alternative HTML version deleted]]
>     >>>     >
>     >>>     > ______________________________________________
>     >>>     > R-help at r-project.org <mailto:R-help at r-project.org>
>     <mailto:R-help at r-project.org <mailto:R-help at r-project.org>>
>     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.
>     >>>     >
>     >>>     >
>     >>>     >
>     >>>     > [[alternative HTML version deleted]]
>     >>>     >
>     >>>     > ______________________________________________
>     >>>     > R-help at r-project.org <mailto:R-help at r-project.org>
>     <mailto:R-help at r-project.org <mailto:R-help at r-project.org>>
>     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.
>     >>>     >
>     >>>     >
>     >>>     >
>     >>>     > Hey,
>     >>>     >
>     >>>     > I have spend some time to write a function, which should
>     >>>     fulfill your needs.
>     >>>     > so i hope ;-)
>     >>>     >
>     >>>     > findIndex<-function(data,element) {
>     >>>     >   ld<-length(data)
>     >>>     >   el<-which(is.element(data,element))
>     >>>     >   lel<-length(el)
>     >>>     >   ndim<-length(dim(data))
>     >>>     >   ind<-array(,dim=c(lel,ndim),dimnames=list(el,1:ndim))
>     >>>     >   precomma<-""
>     >>>     >   tempdata<-data
>     >>>     >   tempel<-el
>     >>>     >   for (j in 1:lel) {
>     >>>     >     data<-tempdata
>     >>>     >     el<-tempel
>     >>>     >     ld<-length(data)
>     >>>     >     for (i in ndim:1) {
>     >>>     >       ratio<-el[j]/(ld/dim(data)[i])
>     >>>     >       if (ratio-trunc(ratio)>0) {
>     >>>     >         ind[j,i]<-trunc(ratio)+1
>     >>>     >       } else {
>     >>>     >         ind[j,i]<-trunc(ratio)
>     >>>     >       }
>     >>>     >       if (length(dim(data))>1) {
>     >>>     >         k<-1
>     >>>     >         while (k>=1 & k<=(i-1)) {
>     >>>     >           precomma<-paste(precomma,",",sep="")
>     >>>     >           k<-k+1
>     >>>     >         }
>     >>>     >
>     >>>     >
>     >>>    
>     data<-as.array(eval(parse(text=paste("data[",precomma,ind[j,i],"]",sep=""))))
>     >>>     >         precomma<-""
>     >>>     >         ld<-length(data)
>     >>>     >         el[j]<-which(is.element(data,element[j]))
>     >>>
>     >>>     >       }
>     >>>     >     }
>     >>>     >   }
>     >>>     >   return(ind)
>     >>>     > }
>     >>>     >
>     >>>     > Regards,
>     >>>     > Christian Porsche
>     >>>     >
>     >>>
>     >>>
>     >>
>     >
>     > ______________________________________________
>     > R-help at r-project.org <mailto:R-help at r-project.org> 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.
>     >
>     >
>
>
>




More information about the R-help mailing list