Hi,
Thank you for the suggestions.
I've attached the code in the text file "test"; to run this one needs the
CollocInfer package(available under R 2.14, now I think if you have 2.14 on
your system you can download CollocInfer from CRAN).
I've tried the 'error' options you suggest; using the syntax
"dimnames(data2)[1:2]" does seem to give an improvement of the attributes of
the array,
But even with the attributes seeming to be in as good order as they might
be(i.e., each boxcar, [[1]],[[2]],[[3]] contains the correct number of names),
rerunning the analysis with the revised array returns the same error.
traceback() gives
4: `colnames<-`(`*tmp*`, value = c("V", "R"))
3: objective(.par, ...)
2: nlminb(coefs, SplineCoefsErr, gradient = SplineCoefsDC, hessian = Hessian,
control = control.in, times = times, data = data, lik = lik,
proc = proc, pars = pars)
1: inneropt(coefs = DEfd2$coefs, times = times, data = data2, lik = lik,
proc = proc, pars = spars, in.meth = "nlminb", control.in = control.out)
To my (untrained) eye, this points again to some kind of mismatch between the
array 'data2' and its dimnames.
Again I'll be grateful if anyone can see what the problem is.
regards,
A
(recover() also is unfruitful)
Fri, 17 Aug 2012 15:28:19 +0000 от William Dunlap <wdun...@tibco.com>:
>
>
>
>
>Have you showed us how to reproduce your original problem?
>
Have you showed us the output of traceback() after encountering
>
the error? Have you tried setting options(error=recover) before
>
encountering the error and then using recover() to look at the dimensions
>
and dimnames of the array that caused the problem?
>
>
This error message
>
> > Error in `colnames<-`(`*tmp*`, value = c("V", "R")) :
>
> > length of 'dimnames' [2] not equal to array extent
>
comes from a nested replacement operation. One such such case is
>
mat <- matrix(1:6,nrow=2,ncol=3)
>
colnames(mat)[1:2] <- c("V", "R")
>
where the nested replacement gets expanded into
>
tmp <- colnames(mat) # tmp becomes NULL
>
tmp[1:2] <- c("V", "R") # tmp becomes c("V","R"), length is 2
>
colnames(tmp) <- tmp # error: 2 colnames for 3 columns
>
(R uses `*tmp*` where I used tmp - the former is for internal use only.)
>
>
There are lots of other possibilities, but you need to at least show the
>
output of traceback() to pin it down.
>
>
Bill Dunlap
>
Spotfire, TIBCO Software
>
wdunlap tibco.com
>
>
>
> -----Original Message-----
>
> From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On
> Behalf
>
> Of aleksandr shfets
>
> Sent: Friday, August 17, 2012 6:25 AM
>
> To: R. Michael Weylandt
>
> Cc: r-help@r-project.org
>
> Subject: Re: [R] dimnames in an array(I'll be grateful if this message will
> be passed to all
>
> list users)
>
>
>
>
>
> Michael,
>
> Thank you for suggestions;
>
> it seems to me that there's a fundamental lacuna with respect to names of a
> three
>
> dimensional array:
>
> that is, rownames fits dimension 1, colnames fits another dimension(that is,
> 3, if I read it
>
> correctly),
>
> but there is no specific name for the third dimension. I've tried "setnames"
> thinking that
>
> this would be appropriate
>
> for naming the sets, but there is no such R function.
>
>
>
> On the other hand you're right, the suggestion
>
> dimnames(data11a)[[2]]=c("V","R")
>
>
>
> doesn't ameliorate the analysis: while now dimnames(data11a)[[2]]
>
>
>
> comes out as
>
> [2]
>
>
>
> returning to the original analysis the same error comes up, that is, that the
> length of
>
> dimnames[2] doesn't agree with thee.
>
> 'array extent' -- that is, even though the boxcar contents are two in number,
> it evaluates
>
> the analysis on the basis of the
>
> dimnames(data11a)[2] value.
>
>
>
> I'm using colloc infer (available under R 2.14) and trying to do the
> FHN(FitzHugh Nagumo)
>
> analysis in section 9.1 of the manual; it's very complex work, but I doubt
> the program is
>
> in error.
>
>
>
> It seems to me in any case that, if R demands that the length of 'dimnames[2]
> must be
>
> equal to the array extent, then there must be a way to make the length of the
> names
>
> equal to the extent of the array, am I right?
>
> Otherwise R would be inconsistent with itself, like Nomad on Startrek, and
> I've never
>
> seen a case where it was thus.
>
>
>
> This is again why I ask if others with extensive experience with dimnames of
> an array
>
> have seen anything similar.
>
> Certainly someone might know what is the significance of the dimnames of the
> array,
>
> how much can the length of them be altered for use generally?
>
>
>
> regards,
>
> A
>
> Thu, 16 Aug 2012 00:00:35 -0400 от "R. Michael Weylandt"
>
> <michael.weyla...@gmail.com>:
>
> >
>
> >
>
> >
>
>
>
>
>
> >
>
>
>
>
>
>
>
> >On Wed, Aug 15, 2012 at 3:57 AM, aleksandr shfets <a_shf...@mail.ru> wrote:
>
> >
>
> > Hi Michael,
>
> >
>
> > Thanks for help on double brackets: I wasn't aware of this use of them.
>
> >
>
> >
>
> >
>
> > I went back to readjust my array so the dimnames(data11a)[[2]] would have
>
> >
>
> > two elements, and remade the array to include the new dimnames:
>
> >
>
> >
>
> >
>
> > dimnames(data11a)[[2]]=c("V","R")
>
> >
>
> >
>
> >
>
> > so that inside the second car of the train would
>
> >
>
> > be two items?
>
> >
>
> >
>
> Yes.
>
> >
>
> >
>
> >
>
> >
>
> > There seems to be no way to define dimnames(data11a)[2] to be two elements
>
> >
>
> > rather than one[as you say we have no way to combine the elements of the
>
> >
>
> > train except using the train].
>
> >
>
> >
>
> You could do this:
>
> >
>
> >
>
> dimnames(data11a)[2] <- list(c("V","R"))
>
> >
>
> >
>
> but I'm not sure that's any clearer. To keep the metaphor rolling (get
>
> >
>
> it!), list() puts your stuff in a boxcar so now you can put it
>
> >
>
> directly on the train.
>
> >
>
> >
>
> > So having tried to do what I can to make the
>
> >
>
> > middle element of the list to two rather than one, I rerun the analysis with
>
> >
>
> > the same result.
>
> >
>
> >
>
> >
>
> > if I read the thing right, regardless of how I define the "inside" of the
>
> >
>
> > boxcar, the analysis that I referred to
>
> >
>
> >
>
> >
>
> > res11a = inneropt(coefs, times=times, data=data11a,
>
> >
>
> > lik=lik,
>
> >
>
> > proc=proc, pars=spars, in.meth='nlminb',
>
> >
>
> > control.in=control.out)
>
> >
>
> >
>
> >
>
> >
>
> Where is this inneropt function from? It could be an error in that. If
>
> >
>
> it's from a package, give me the name and I'll take a look at it and
>
> >
>
> you can also contact the maintainer: get contact info using the
>
> >
>
> maintainer() function.
>
> >
>
> >
>
> Also, some of the advice here will probably be helpful in seeking follow-up:
>
> >
>
> >http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-
>
> example
>
> >
>
> >
>
> Cheers,
>
> >
>
> Michael
>
> >
>
> >
>
> > still won't look at the 'contents' of the car, but gives the same message:
>
> >
>
> >
>
> >
>
> > Error in `colnames<-`(`*tmp*`, value = c("V", "R")) :
>
> >
>
> > length of 'dimnames' [2] not equal to array extent
>
> >
>
> >
>
> >
>
> >
>
> >
>
> > I've tried a couple of other things like, after defining the second boxcar
>
> >
>
> > as:
>
> >
>
> >
>
> >
>
> > dimnames(data11a)[[2]]=c("V","R")
>
> >
>
> >
>
> >
>
> >
>
> >
>
> > then
>
> >
>
> >
>
> >
>
> >
>
> >
>
> > dimnames(data11a)[2]=list(dimnames(data11a)[[2]])
>
> >
>
> >
>
> >
>
> >
>
> >
>
> > but this last generates an error.
>
> >
>
> >
>
> >
>
> > Somehow I think that that the clue to the thing is in the error message:
>
> >
>
> > what I see[ with my untrained eye] is that there should be some way to
>
> >
>
> > reconcile the three-- colnames, dimnames and the array- to a greater degree
>
> >
>
> > than their present agreement.
>
> >
>
> > Am I right?
>
> >
>
> > regards,
>
> >
>
> > Russell
>
> >
>
> >
>
> >
>
> > Tue, 14 Aug 2012 01:27:50 -0400 от "R. Michael Weylandt"
>
> >
>
> > <michael.weyla...@gmail.com>:
>
> >
>
> >
>
> >
>
> > On Mon, Aug 13, 2012 at 5:10 AM, aleksandr russell <sss...@gmail.com> wrote:
>
> >
>
> >> Hello,
>
> >
>
> >>
>
> >
>
> >> I'm hoping someone with a wide experience with R may be able to see
>
> >
>
> >> what the program is trying to tell me.
>
> >
>
> >>
>
> >
>
> >> I've got an array:
>
> >
>
> >>
>
> >
>
> >> y1=rnorm(41,0.2)
>
> >
>
> >> y2=rnorm(41,0.2)
>
> >
>
> >> y3=rbind(y1,y2)
>
> >
>
> >>
>
> >
>
> >>
>
> >
>
> >>
>
> >
>
> >> data11<-array(0,c(41,2,2))
>
> >
>
> >> data11[,1,]=y3
>
> >
>
> >> data11[,2,]=y3
>
> >
>
> >> rownames(data11)<-rownames(data11, do.NULL = FALSE, prefix = "Obs.")
>
> >
>
> >> colnames=c("V","R")
>
> >
>
> >> varnames=c("one","two")
>
> >
>
> >> dimnames(data11)<-list(rownames(data11), varnames, colnames)
>
> >
>
> >>
>
> >
>
> >>
>
> >
>
> >>
>
> >
>
> >> data11a<-as.array(data11, dimnames=dimnames(data11))
>
> >
>
> >
>
> >
>
> > # No effect here
>
> >
>
> > identical(data11, data11a) # TRUE
>
> >
>
> >
>
> >
>
> >>
>
> >
>
> >> The analysis that I would like to do with R(CollocInfer) runs as
>
> >
>
> >> follows(I print for the sake of introduction here though to run this
>
> >
>
> >> line of code requires extensive setup)
>
> >
>
> >>
>
> >
>
> >> res11a = inneropt(coefs, times=times, data=data11a, lik=lik,
>
> >
>
> >> proc=proc, pars=spars, in.meth='nlminb', control.in=control.out)
>
> >
>
> >>
>
> >
>
> >>
>
> >
>
> >>
>
> >
>
> >> The response I get is
>
> >
>
> >>
>
> >
>
> >> Error in 'colnames<-'('*tmp*', value = c("V","R")) :
>
> >
>
> >> length of 'dimnames' [2] not equal to array extent
>
> >
>
> >>
>
> >
>
> >>
>
> >
>
> >>
>
> >
>
> >> It seems to me that there is a fundamental match of 'dimnames' with
>
> >
>
> >> the array that is being referred to.
>
> >
>
> >>
>
> >
>
> >> I have learned that the normal length of dimnames for an array is one
>
> >
>
> >> for each element in the list: thus as I've set up the array, the
>
> >
>
> >> length(dimnames(data11a)[2]) gives me
>
> >
>
> >>
>
> >
>
> >> [1]
>
> >
>
> >
>
> >
>
> > I believe you are looking for
>
> >
>
> >
>
> >
>
> > length(dimnames(data11a)[[2]]) # 2
>
> >
>
> >
>
> >
>
> > Look into the difference between `[` and `[[` for subsetting a list.
>
> >
>
> > See ?Extract for the gorey details, but I like my little train
>
> >
>
> > metaphor:
>
> >
>
> >
>
> >
>
> > If the list "x", is a train: x[2] is the "sub-train" consisting only
>
> >
>
> > of the second car, while x[[2]] is the contents of that second car.
>
> >
>
> > Hence we can do x[1:3] to give a well defined train, but not x[[1:3]]
>
> >
>
> > because we have no way to combine the elements of the first three cars
>
> >
>
> > other than using a train.
>
> >
>
> >
>
> >
>
> > Cheers,
>
> >
>
> > Michael
>
> >
>
> >
>
> >
>
> >>
>
> >
>
> >> I would like to ask :Is it not evident from R's response that there is
>
> >
>
> >> some other possibility for the dimnames[2] of the array?
>
> >
>
> >>
>
> >
>
> >> If so, what is it?
>
> >
>
> >>
>
> >
>
> >> regards,
>
> >
>
> >> A
>
> >
>
> >>
>
> >
>
> >> ______________________________________________
>
> >
>
> >> R-help@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]]
>
>
times = seq(0,20,0.5)
pars = c(0.2,0.2,3)
names(pars) = c('a','b','c')
x0 = c(-1,1)
names(x0) = c('V','R')
fhn = make.fhn()
y = lsoda(x0,times,fhn$fn.ode,pars)
y = y[,2:3]
w=rnorm(41,.05)
z=rnorm(41,.05)
g=y+w
assign("y", lsoda(x0,times,fhn$fn.ode,pars)
)
y=y[,2:3]
knots = seq(0,20,0.2)
norder = 3
nbasis = length(knots) + norder - 2
range = c(0,20)
bbasis = create.bspline.basis(range=range, nbasis=nbasis, norder=norder,
breaks=knots)
fd.data = array(data,c(nrow(data),1,ncol(data)))
m2<- cbind(.29, 1:41)
m3=m2[,1]
m4<-rbind(m3,m3)
m6<- cbind(.29, 1:82)
m7<-as.array(m6,dim=c(41,2))
m8<-cbind(.29,1:41)
m9<-as.array(m8,dim=c(41,2))
data<-y+m9
fd.data = array(data,c(nrow(data),1,ncol(data)))
fd.data = array(data,c(nrow(data),1,ncol(data)))
varnames=c("V","R")
DEfd = Data2fd(fd.data, times, bbasis,
fdnames=list(NULL,NULL,varnames) )
coefs = DEfd$coefs
colnames(coefs) = varnames
control=list()
control$trace = 0
control$maxit = 1000
control$maxtry = 10
control$reltol = 1e-6
control$meth = "BFGS"
control.in = control
control.in$reltol = 1e-12
control.in$print.level = 0
control.in$iterlim = 1000
control.out = control
control.out$trace = 2
lambda = c(10000,10000)
res0 = Profile.LS(fhn,data,times,pars,coefs=coefs,basisvals=bbasis,
lambda=lambda,in.meth='nlminb',out.meth='ProfileGN',
control.in=control.in,control.out=control.out)
res1 = Profile.LS(fhn$fn, data, times, pars=spars, coefs=coefs,
lambda=lambda, in.meth='nlminb', out.meth='nls',
control.in=control.in, control.out=control.out)
spars=c(0.3,0.1,2)
res1 = Profile.LS(fhn$fn, data, times, pars=spars, coefs=coefs,
lambda=lambda, in.meth='nlminb', out.meth='nls',
control.in=control.in, control.out=control.out)
profile.obj = LS.setup(pars=pars, fn=fhn, basisvals=bbasis,
lambda=10000, times=times, fd.obj=DEfd)
lik = profile.obj$lik
proc = profile.obj$proc
qpts = knots
qwts = rep(1/length(knots),length(knots))
qwts = qwts %*% t(lambda)
weights = array(1,dim(data))
likmore = make.id()
likmore$weights = weights
lik = make.SSElik()
lik$more = likmore
lik$bvals = eval.basis(times,bbasis)
procmore = make.fhn()
procmore$weights = qwts
procmore$qpts = qpts
procmore$names = varnames
parnames=c("a","b","c")
procmore$parnames = parnames
procmore$parnames = parnames
proc = make.SSEproc()
proc$more = procmore
proc$bvals = list(bvals = eval.basis(procmore$qpts,bbasis,0),
dbvals = eval.basis(procmore$qpts,bbasis,1))
spars = c(0.3,0.1,2)
res0 = inneropt(coefs, times=times, data=data, lik=lik, proc=proc,
pars=spars, in.meth='nlminb', control.in=control.out)
ncoefs = res0$coefs
res1 = outeropt(data=data, times=times, pars=pars, coefs=coefs,
lik=lik, proc=proc, in.meth="nlminb", out.meth="nls",
control.in=control.in, control.out=control.out)
res2 = outeropt(data=data, times=times, pars=pars, coefs=coefs,
lik=lik, proc=proc,
in.meth="nlminb", out.meth="nlminb",
control.in=control.in, control.out=control.out)
Profile.covariance(pars=res1$pars, times=times, data=data,
coefs=res1$coefs, lik=lik, proc=proc)
data2 = array(0,c(41,2,2))
dim(y)
dim(w)
m2<- cbind(.29, 1:41)
m3=m2[,1]
m4<-rbind(m3,m3)
m6<- cbind(.29, 1:82)
m7<-as.array(m6,dim=c(41,2))
m8<-cbind(.29,1:41)
m9<-as.array(m8,dim=c(41,2))
data<-y+m9
data2[,1,] = y+m9
data2[,2,] = y+m9
fd.data2 = array(data2,c(nrow(data2),2,ncol(data2)))
DEfd2 = Data2fd(fd.data2, times, bbasis,
fdnames=list(NULL,NULL,varnames))
rownames(data2)<-rownames(data2, do.NULL = FALSE, prefix = "Obs.")
colnames(data2)[1:2]=c("one","two")
varnames(data2)[1:2]=c("V","R")
dimnames(data2)=list(rownames(data2),colnames(data2),varnames)
data2a<-as.array(data2,dim=c(41,2,2),dimnames=dimnames(data2))
profile2.obj=LS.setup(pars,DEfd2$coefs,fhn,basisvals=bbasis,lambda,fd.obj=NULL,
more=NULL,data=data2,weights=NULL,times=times,quadrature=NULL,eps=1e-6,
posproc=FALSE,poslik=FALSE,discrete=FALSE,names=NULL,sparse=FALSE)
assign("proc", profile2.obj$proc)
qpts = knots
qwts = rep(1/length(knots),length(knots))
qwts = qwts %*% t(lambda)
weights = array(1,dim(data2))
likmore = make.id()
likmore$weights = weights
lik = make.SSElik()
lik$more = likmore
lik$bvals = eval.basis(times,bbasis)
procmore = make.fhn()
procmore$weights = qwts
procmore$qpts = qpts
procmore$names = varnames
procmore$parnames = parnames
proc = make.SSEproc()
proc$more = procmore
proc$bvals = list(bvals = eval.basis(procmore$qpts,bbasis,0),
dbvals = eval.basis(procmore$qpts,bbasis,1))
res10 = inneropt(coefs=DEfd2$coefs, times=times, data=data2, lik=lik,
proc=proc,
pars=spars, in.meth='nlminb', control.in=control.out)
______________________________________________
R-help@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.
