CRAN Package Check Results for Package rarhsmm

Last updated on 2019-05-01 01:51:59 CEST.

Check Details

Version: 1.0.7
Check: examples
Result: ERROR
    Running examples in 'rarhsmm-Ex.R' failed
    The error most likely occurred in:
    
    > base::assign(".ptime", proc.time(), pos = "CheckExEnv")
    > ### Name: em.hmm
    > ### Title: EM algorithm to compute maximum likelihood estimate of Gaussian
    > ### hidden Markov models with / without autoregressive structures and
    > ### with / without regularization on the covariance matrices and/or
    > ### autoregressive structures.
    > ### Aliases: em.hmm
    >
    > ### ** Examples
    >
    > set.seed(332213)
    > data(finance)
    > x <- data.matrix(finance)
    > #log return
    > y <- x[-1,-51]
    > for(i in 2:nrow(x)){
    + y[i-1,] <- log(x[i,-51]) - log(x[i-1,-51])
    + }
    > #annualize the log return
    > y <- y * 252
    >
    > #first, fit a Gaussian HMM without autoregressive structure
    > m <- 2
    > #initialize the list of means
    > mu <- list(apply(y,2,mean), apply(y,2,mean))
    > #initialize the list of covariance matrices
    > sigma <- list(cov(y)*1.2,cov(y)*0.8)
    > #initialize the prior probability
    > delta <- c(0.5,0.5)
    > #initialize the transition probabilities
    > gamma <- matrix(c(0.9,0.1,0.2,0.8),2,2,byrow=TRUE)
    > mod1 <- list(m=m,mu=mu,sigma=sigma,delta=delta,gamma=gamma)
    > #will not run without a shrinkage on the covariance matrices because the
    > #series is not long enough to reliably estimate the covariance structure
    > fit1 <- em.hmm(y=y,mod=mod1,cov.shrink=0.0001)
    Error in getnodeprob_nocov_mvn(y, mu, sigma, K, p, 0, 0) :
     BLAS/LAPACK routine 'DTRTI2' gave error code -1
    Calls: em.hmm -> em.mvn -> getnodeprob_nocov_mvn
    Execution halted
Flavor: r-devel-linux-x86_64-debian-clang

Version: 1.0.7
Check: examples
Result: ERROR
    Running examples in ‘rarhsmm-Ex.R’ failed
    The error most likely occurred in:
    
    > base::assign(".ptime", proc.time(), pos = "CheckExEnv")
    > ### Name: em.hmm
    > ### Title: EM algorithm to compute maximum likelihood estimate of Gaussian
    > ### hidden Markov models with / without autoregressive structures and
    > ### with / without regularization on the covariance matrices and/or
    > ### autoregressive structures.
    > ### Aliases: em.hmm
    >
    > ### ** Examples
    >
    > set.seed(332213)
    > data(finance)
    > x <- data.matrix(finance)
    > #log return
    > y <- x[-1,-51]
    > for(i in 2:nrow(x)){
    + y[i-1,] <- log(x[i,-51]) - log(x[i-1,-51])
    + }
    > #annualize the log return
    > y <- y * 252
    >
    > #first, fit a Gaussian HMM without autoregressive structure
    > m <- 2
    > #initialize the list of means
    > mu <- list(apply(y,2,mean), apply(y,2,mean))
    > #initialize the list of covariance matrices
    > sigma <- list(cov(y)*1.2,cov(y)*0.8)
    > #initialize the prior probability
    > delta <- c(0.5,0.5)
    > #initialize the transition probabilities
    > gamma <- matrix(c(0.9,0.1,0.2,0.8),2,2,byrow=TRUE)
    > mod1 <- list(m=m,mu=mu,sigma=sigma,delta=delta,gamma=gamma)
    > #will not run without a shrinkage on the covariance matrices because the
    > #series is not long enough to reliably estimate the covariance structure
    > fit1 <- em.hmm(y=y,mod=mod1,cov.shrink=0.0001)
    
     *** caught segfault ***
    address 0x9, cause 'memory not mapped'
    
    Traceback:
     1: getnodeprob_nocov_mvn(y, mu, sigma, K, p, 0, 0)
     2: em.mvn(y, mod, ntimes, tol, maxit, cov.shrink, print)
     3: em.hmm(y = y, mod = mod1, cov.shrink = 1e-04)
    An irrecoverable exception occurred. R is aborting now ...
    Segmentation fault
Flavor: r-release-linux-x86_64