[R-sig-ME] Out of memory with spatial correlation.

[Doran, Harold]

Thierry

One option is to switch to lmer rather than lme. Of course, with lmer
you don't have the same options for correlation structures. So, what you
might consider is to "pre-whiten" your data using the methods outlined
in Pinhiero and Bates and then run the model using lmer.

In fact, that is what lme does. It "pre-whitens" the data and then runs
the model under normal assumptions. The only difference here is that you
would need to pre-whiten manually. 

> -----Original Message-----
> From: r-sig-mixed-models-bounces at r-project.org 
> [mailto:r-sig-mixed-models-bounces at r-project.org] On Behalf 
> Of ONKELINX, Thierry
> Sent: Tuesday, December 09, 2008 7:59 AM
> To: R-sig-mixed-models at r-project.org
> Subject: [R-sig-ME] Out of memory with spatial correlation.
> 
> Dear all,
> 
> My model runs fine with just the fixed and random effects. 
> But when I add a spatial correlation structure it runs out of 
> memory (Error: cannot allocate vector of size 185.6 Mb). The 
> problem is that the data is clearly spatially correlated. I 
> have tried the simplify the fixed effects and the random 
> effects with no avail. So the problem is probably in the 
> correlation structure (see code below). Any suggestion on how 
> to incorporate the spatial autocorrelation?
> 
> A description of the design.
> 
> We are testing a methodology to monitor bats. Basically 
> volunteers ride along a predefined route by car (30 km/h) or 
> by bike (15 km/h). They record the echolocation sounds of the 
> bats at fixed intervals. There position is tracked by GPS so 
> we know were each recording was made. An expert counts the 
> number of pulses in each recording.
> We predefined 10 routes, 5 for the cars and 5 for the bikes. 
> Each car route overlaps with one bike route. So we have 5 
> groups of routes.
> Within each group we have an longer car route that overlaps 
> with a shorter bike route. There is no overlap between 
> groups. All routes were driven three times, all vehicules 
> started their route simultanious.
> We have about 400 recordings per route of 10 routes at 3 
> occasions = 12000 rows.
> 
> Our nullhypothese is that the average number of recordings 
> does not depend on the type of vehicule.
> 
> The model that fails is:
> library(nlme)
> lme(log(Pulses + 1) ~ Transport + Occasion, random = ~ 1| 
> Group, correlation = corExp(form = ~ X + Y))
> 
> The model works if I omit the correlation structure:
> lme(log(Pulses + 1) ~ Transport + Occasion, random = ~ 1| Group)
> 
> I'm using R 2.8.0 with nlme 3.1-89 on WinXP with 2GB RAM. 
> --mem-size is set at the maximum (2047 MB).
> 
> Regards,
> 
> Thierry
> 
> 
> --------------------------------------------------------------
> ----------
> ----
> ir. Thierry Onkelinx
> Instituut voor natuur- en bosonderzoek / Research Institute 
> for Nature and Forest Cel biometrie, methodologie en 
> kwaliteitszorg / Section biometrics, methodology and quality 
> assurance Gaverstraat 4 9500 Geraardsbergen Belgium tel. + 32 
> 54/436 185 Thierry.Onkelinx at inbo.be www.inbo.be 
> 
> To call in the statistician after the experiment is done may 
> be no more than asking him to perform a post-mortem 
> examination: he may be able to say what the experiment died of.
> ~ Sir Ronald Aylmer Fisher
> 
> The plural of anecdote is not data.
> ~ Roger Brinner
> 
> The combination of some data and an aching desire for an 
> answer does not ensure that a reasonable answer can be 
> extracted from a given body of data.
> ~ John Tukey
> 
> Dit bericht en eventuele bijlagen geven enkel de visie van de 
> schrijver weer en binden het INBO onder geen enkel beding, 
> zolang dit bericht niet bevestigd is door een geldig 
> ondertekend document. The views expressed in  this message 
> and any annex are purely those of the writer and may not be 
> regarded as stating an official position of INBO, as long as 
> the message is not confirmed by a duly signed document.
> 
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list 
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>