[R] Segmentation fault in estimating structural equation models with the SEM package.

Snaebjorn Gunnsteinsson sgunnste at jhsph.edu
Fri Mar 16 13:03:40 CET 2007


Dear R-users,

     I am running a large number of simulations and estimating a  
structural equation model for each one using the SEM package. Each  
run of my program has around 8000 simulations. Most of the time the  
program completes all of them correctly but sometimes I get a  
segmentation fault in the sem routine and my program stops with the  
following error message:

>  *** caught segfault ***
> address (nil), cause 'unknown'
>
> Traceback:	
> 1: nlm(if (analytic.gradient) objective.2 else objective.1,  
> start,     hessian = TRUE, iterlim = maxiter, print.level = if  
> (debug) 2 else 0,     typsize = typsize, .
> ..)
> 2: sem.default(ram = ram, S = S, N = N, param.names = pars,  
> var.names = vars,     fixed.x = fixed.x, debug = debug, ...)
> 3: sem(ram = ram, S = S, N = N, param.names = pars, var.names =  
> vars,     fixed.x = fixed.x, debug = debug, ...)
> 4: sem.mod(modelspec, cor.or.cov.mat, N, obs.variables =  
> comb.set,     warn = FALSE)
> 5: sem(modelspec, cor.or.cov.mat, N, obs.variables = comb.set, warn  
> = FALSE)
> 6: try(sem(modelspec, cor.or.cov.mat, N, obs.variables =  
> comb.set,     warn = FALSE), silent = TRUE)
> 7: eval.with.vis(expr, envir, enclos)
> 8: eval.with.vis(ei, envir)
> 9: source("analysis.base.R")
> aborting ...

Does anybody have an idea about how I could prevent this from  
happening or catch it so that my program could continue running?  
Could I, in some way, manage the memory myself to prevent this?

The model is correctly specified and clearly identified but the  
covariance matrix changes between simulations (it is a two latent  
variable measurement model with more than 4 observed variables on  
each LV).

Many thanks,
regards,
Snaebjorn

P.S. Using R 2.4.1 (2006-06-01) on x86_64 GNU/Linux
Using version 0.9-6 (2006/11/22) of the SEM package.


-----------------------------------------------------------
Snaebjorn Gunnsteinsson
JiVitA Science Fellow
The JiVitA Project (NIPHP-USAID)
Rangpur, Bangladesh
www.jivita.org
Mobile: +880-171-320-2560



More information about the R-help mailing list