[R] sample variogram construction

[Dan Bebber]

Matt,

In this case you are plotting the variogram of the residuals of the lme
object, not of the data themselves. In your model you are assuming a linear
relationship between count and time, with different intercepts and slopes
for your different individuals. You are also assuming that the residuals
exhibit stationarity. The Variogram function *should* work, so I can only
suggest that there is something wrong with your data or code. "Mixed-Effects
Models in S and S-Plus" by Pinheiro and Bates will probably help.

Dan

Department of Plant Sciences
University of Oxford
South Parks Road
Oxford OX1 3RB
UK
Tel. 01865 275000



> -----Original Message-----
> From: matt.kowgier at utoronto.ca [mailto:matt.kowgier at utoronto.ca] 
> Sent: 25 October 2004 18:32
> To: danbebber at forestecology.co.uk; r-help at stat.math.ethz.ch
> Subject: RE: [R] sample variogram construction
> 
> 
> Okay thanks!
> No I have the following difficulty implementing a variogram 
> in the nlme package:
> 
> > cd1 <- lme(count ~ time, data=cd4,random= ~ time | id) 
> > plot(Variogram(cd1, form= ~ time | id, robust=TRUE))
> Error in as.array(X) : attempt to set an attribute on NULL
> 
> Im not sure how to fix this, and apply this function to
> a longitudinal data with unequal times?
> 
> Any help would be appreciated
> 
> 
> Quoting Dan Bebber <danbebber at forestecology.co.uk>:
> 
> > Hi Matt,
> > 
> > there are several R packages that will compute the sample variogram 
> > for you. Check out GeoR, sgeostat, nlme, spatial. There's 
> no point in 
> > recoding the whole lot yourself, unless as a learning excercise.
> > 
> > D
> > 
> > p.s. For time series autocorrelations, you could use acf in package 
> > stats.
> > 
> > Message: 9
> > Date: Mon, 25 Oct 2004 02:02:06 -0400
> > From: matt.kowgier at utoronto.ca
> > Subject: [R] sample variogram construction
> > To: r-help at stat.math.ethz.ch
> > Message-ID: <1098684126.417c96de334e1 at webmail.utoronto.ca>
> > Content-Type: text/plain; charset=US-ASCII
> > 
> > Hi
> > 
> > Im attempting to build a sample variogram for 300 obersvations of 
> > longitudinal data. So what I need to do is compute the half squared 
> > differences  between pairs of residuals (for instance if a 
> subject has 
> > 4 obersvations, this is 4 choose 2 paird differences) for each 
> > subject. Also, then I need the corresponding time 
> differences within 
> > each individual. So the end result will be a 300 by 2 matrix with 
> > columns corresponding to paired difference residuals within subject 
> > and time differences within subject. Basically im having trouble 
> > coding this kind of matrix in R, if anyone can help me out 
> or give me 
> > some tips id appreciate it.
> > 
> > Thanks.
> > Stuck in the for loop
> > student
> > 
> > 
> 
>