CRAN Package Check Results for Package pan

Last updated on 2017-04-28 23:52:33.

Flavor Version Tinstall Tcheck Ttotal Status Flags
r-devel-linux-x86_64-debian-clang 1.4 3.77 19.49 23.26 NOTE
r-devel-linux-x86_64-debian-gcc 1.4 4.18 18.91 23.09 NOTE
r-devel-linux-x86_64-fedora-clang 1.4 41.80 NOTE --no-stop-on-test-error
r-devel-linux-x86_64-fedora-gcc 1.4 39.04 NOTE --no-stop-on-test-error
r-devel-windows-ix86+x86_64 1.4 13.00 70.00 83.00 OK
r-patched-linux-x86_64 1.4 3.89 18.52 22.41 NOTE
r-patched-solaris-sparc 1.4 224.40 OK
r-patched-solaris-x86 1.4 44.20 ERROR
r-release-linux-x86_64 1.4 3.69 18.93 22.62 NOTE
r-release-windows-ix86+x86_64 1.4 18.00 53.00 71.00 OK
r-release-osx-x86_64 1.4 WARN
r-oldrel-windows-ix86+x86_64 1.4 17.00 43.00 60.00 OK
r-oldrel-osx-x86_64 1.4 OK

Check Details

Version: 1.4
Check: compiled code
Result: NOTE
    File ‘pan/libs/pan.so’:
     Found no calls to: ‘R_registerRoutines’, ‘R_useDynamicSymbols’
    
    It is good practice to register native routines and to disable symbol
    search.
    
    See ‘Writing portable packages’ in the ‘Writing R Extensions’ manual.
Flavors: r-devel-linux-x86_64-debian-clang, r-devel-linux-x86_64-debian-gcc, r-patched-linux-x86_64, r-release-linux-x86_64

Version: 1.4
Flags: --no-stop-on-test-error
Check: compiled code
Result: NOTE
    File ‘pan/libs/pan.so’:
     Found no calls to: ‘R_registerRoutines’, ‘R_useDynamicSymbols’
    
    It is good practice to register native routines and to disable symbol
    search.
    
    See ‘Writing portable packages’ in the ‘Writing R Extensions’ manual.
Flavors: r-devel-linux-x86_64-fedora-clang, r-devel-linux-x86_64-fedora-gcc

Version: 1.4
Check: examples
Result: ERROR
    Running examples in ‘pan-Ex.R’ failed
    The error most likely occurred in:
    
    > ### Name: pan
    > ### Title: Imputation of multivariate panel or cluster data
    > ### Aliases: pan
    > ### Keywords: models
    >
    > ### ** Examples
    >
    > ########################################################################
    > # This example is somewhat atypical because the data consist of a
    > # single response variable (change in heart rate) measured repeatedly;
    > # most uses of pan() will involve r > 1 response variables. If we had
    > # r response variables rather than one, the only difference would be
    > # that the vector y below would become a matrix with r columns, one
    > # for each response variable. The dimensions of Sigma (the residual
    > # covariance matrix for the response) and Psi (the covariance matrix
    > # for the random effects) would also change to (r x r) and (r*q x r*q),
    > # respectively, where q is the number of random coefficients in the
    > # model (in this case q=1 because we have only random intercepts). The
    > # new dimensions for Sigma and Psi will be reflected in the prior
    > # distribution, as Dinv and Binv become (r x r) and (r*q x r*q).
    > #
    > # The pred matrix has the same number of rows as y, the number of
    > # subject-occasions. Each column of Xi and Zi must be represented in
    > # pred. Because Zi is merely the first column of Xi, we do not need to
    > # enter that column twice. So pred is simply the matrix Xi, stacked
    > # upon itself nine times.
    > #
    > data(marijuana)
    > attach(marijuana)
    > pred <- with(marijuana,cbind(int,dummy1,dummy2,dummy3,dummy4,dummy5))
    > #
    > # Now we must tell pan that all six columns of pred are to be used in
    > # Xi, but only the first column of pred appears in Zi.
    > #
    > xcol <- 1:6
    > zcol <- 1
    > ########################################################################
    > # The model specification is now complete. The only task that remains
    > # is to specify the prior distributions for the covariance matrices
    > # Sigma and Psi.
    > #
    > # Recall that the dimension of Sigma is (r x r) where r
    > # is the number of response variables (in this case, r=1). The prior
    > # distribution for Sigma is inverted Wishart with hyperparameters a
    > # (scalar) and Binv (r x r), where a is the imaginary degrees of freedom
    > # and Binv/a is the prior guesstimate of Sigma. The value of a must be
    > # greater than or equal to r. The "least informative" prior possible
    > # would have a=r, so here we will take a=1. As a prior guesstimate of
    > # Sigma we will use the (r x r) identity matrix, so Binv = 1*1 = 1.
    > #
    > # By similar reasoning we choose the prior distribution for Psi. The
    > # dimension of Psi is (r*q x r*q) where q is the number of random
    > # effects in the model (i.e. the length of zcol, which in this case is
    > # one). The hyperparameters for Psi are c and Dinv, where c is the
    > # imaginary degrees of freedom (which must be greater than or equal to
    > # r*q) and Dinv/c is the prior guesstimate of Psi. We will take c=1
    > # and Dinv=1*1 = 1.
    > #
    > # The prior is specified as a list with four components named a, Binv,
    > # c, and Dinv, respectively.
    > #
    > prior <- list(a=1,Binv=1,c=1,Dinv=1)
    > ########################################################################
    > # Now we are ready to run pan(). Let's assume that the pan function
    > # and the object code have already been loaded into R. First we
    > # do a preliminary run of 1000 iterations.
    > #
    > result <- pan(y,subj,pred,xcol,zcol,prior,seed=13579,iter=1000)
    > #
    > # Check the convergence behavior by making time-series plots and acfs
    > # for the model parameters. Variances will be plotted on a log
    > # scale. We'll assume that a graphics device has already been opened.
    > #
    > plot(1:1000,log(result$sigma[1,1,]),type="l")
    > acf(log(result$sigma[1,1,]))
    > plot(1:1000,log(result$psi[1,1,]),type="l")
    > acf(log(result$psi[1,1,]))
    > par(mfrow=c(3,2))
    > for(i in 1:6) plot(1:1000,result$beta[i,1,],type="l")
    > for(i in 1:6) acf(result$beta[i,1,])
    > #
    > # This example appears to converge very rapidly; the only appreciable
    > # autocorrelations are found in Psi, and even those die down by lag
    > # 10. With a sample this small we can afford to be cautious, so let's
    > # impute the missing data m=10 times taking 100 steps between
    > # imputations. We'll use the current simulated value of y as the first
    > # imputation, then restart the chain where we left off to produce
    > # the second through the tenth.
    > #
    > y1 <- result$y
    > result <- pan(y,subj,pred,xcol,zcol,prior,seed=9565,iter=100,start=result$last)
    > y2 <- result$y
    > result <- pan(y,subj,pred,xcol,zcol,prior,seed=6047,iter=100,start=result$last)
    > y3 <- result$y
    > result <- pan(y,subj,pred,xcol,zcol,prior,seed=3955,iter=100,start=result$last)
    > y4 <- result$y
    > result <- pan(y,subj,pred,xcol,zcol,prior,seed=4761,iter=100,start=result$last)
    > y5 <- result$y
    > result <- pan(y,subj,pred,xcol,zcol,prior,seed=9188,iter=100,start=result$last)
    Error in pan(y, subj, pred, xcol, zcol, prior, seed = 9188, iter = 100, :
     NA/NaN/Inf in foreign function call (arg 20)
    Execution halted
Flavor: r-patched-solaris-x86

Version: 1.4
Check: re-building of vignette outputs
Result: WARN
    Error in re-building vignettes:
     ...
    Error in texi2dvi(file = file, pdf = TRUE, clean = clean, quiet = quiet, :
     Running 'texi2dvi' on 'pan-tr.tex' failed.
    LaTeX errors:
    ! LaTeX Error: File `a4wide.sty' not found.
    
    Type X to quit or <RETURN> to proceed,
    or enter new name. (Default extension: sty)
    
    ! Emergency stop.
    <read *>
    
    l.8 \usepackage
     {amsmath,epsfig}^^M
    ! ==> Fatal error occurred, no output PDF file produced!
    Calls: buildVignettes -> texi2pdf -> texi2dvi
    Execution halted
Flavor: r-release-osx-x86_64