[R] error in arfima...
Martin Maechler
m@ech|er @end|ng |rom @t@t@m@th@ethz@ch
Mon Jun 5 11:49:20 CEST 2023
> Dear Martin,
> REgrets to reply this late....
> I am staring at a conundrum never before encountered in my experience with R:
> LYGH[[201]]
> [1] 45.40 3.25 6.50 2.15
> > arfima(LYGH[[201]])
> Error in .fdcov(x, fdf$d, h, nar = nar, nma = nma, hess = hess, fdf.work = fdf$w) :
> NA/NaN/Inf in foreign function call (arg 5)
> > arfima(c(45.40,3.25,6.50,2.15))
> Call:
> arfima(y = c(45.4, 3.25, 6.5, 2.15))
> Coefficients:
> d
> 4.583013e-05
> sigma[eps] = 18.01252
> a list with components:
> [1] "log.likelihood" "n" "msg" "d" "ar"
> [6] "ma" "covariance.dpq" "fnormMin" "sigma" "stderror.dpq"
> [11] "correlation.dpq" "h" "d.tol" "M" "hessian.dpq"
> [16] "length.w" "residuals" "fitted" "call" "x"
> [21] "series"
> Please note that the index of LYGH has changed from 202 to 201 due to some randomness in one of my function.
> PLEASE HELP.
> Output of dput LYGH[[201]]:
> > dput(LYGH[[201]])
> c(45.4, 3.25, 6.5, 2.14999999999998)
> output of session info()
> sessionInfo()
> R version 4.1.2 (2021-11-01)
> Platform: x86_64-w64-mingw32/x64 (64-bit)
> Running under: Windows Server x64 (build 14393)
> Matrix products: default
> locale:
> [1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252
> [3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C
> [5] LC_TIME=English_United States.1252
> attached base packages:
> [1] parallel stats graphics grDevices utils datasets methods base
> other attached packages:
> [1] pbmcapply_1.5.1 imputeTS_3.3 forecast_8.17.0
> loaded via a namespace (and not attached):
> [1] Rcpp_1.0.7 urca_1.3-3 pillar_1.9.0 compiler_4.1.2 tseries_0.10-51
> [6] tools_4.1.2 xts_0.12.1 nlme_3.1-153 lifecycle_1.0.3 tibble_3.2.1
> [11] gtable_0.3.3 lattice_0.20-45 pkgconfig_2.0.3 rlang_1.1.0 cli_3.6.1
> [16] rstudioapi_0.14 curl_4.3.2 xml2_1.3.3 dplyr_1.1.1 generics_0.1.3
> [21] vctrs_0.6.1 gridtext_0.1.5 ggtext_0.1.2 lmtest_0.9-40 grid_4.1.2
> [26] nnet_7.3-16 tidyselect_1.2.0 glue_1.6.2 R6_2.5.1 fansi_1.0.4
> [31] ggplot2_3.4.2 TTR_0.24.3 magrittr_2.0.3 scales_1.2.1 quantmod_0.4.20
> [36] timeDate_4021.106 colorspace_2.1-0 fracdiff_1.5-1 quadprog_1.5-8 utf8_1.2.3
> [41] stinepack_1.4 munsell_0.5.0 zoo_1.8-10
> PLease Also note:
> arfima(c(45.4, 3.25, 6.5, 2.14999999999998))
> Call:
> arfima(y = c(45.4, 3.25, 6.5, 2.14999999999998))
> Coefficients:
> d
> 4.583013e-05
> sigma[eps] = 18.01252
> a list with components:
> [1] "log.likelihood" "n" "msg" "d" "ar"
> [6] "ma" "covariance.dpq" "fnormMin" "sigma" "stderror.dpq"
> [11] "correlation.dpq" "h" "d.tol" "M" "hessian.dpq"
> [16] "length.w" "residuals" "fitted" "call" "x"
> [21] "series"
> Many thanks in advance....
> Thanking you,
> Yours sincerely,
> AKSHAY M KULKARNI
Hmm... indeed, "conundrum" may be a euphemism for the
situation..... which clearly points to a bug *somewhere*.
As your dput() shows, the argument *is* a simple vector .. and
even if the default dput() may loose a few bits in
precison... you did show that with (possibly only almost)
identical argument, the error did not happen.
I really have no idea what's going on, ....
Some musing/ideas: the NA/NaN/Inf warning
may come from a memory corruption problem somewhere because some
of the involved packages using compiled code or the old version of R/lapack/???
or ??? cause it.
Does the error only happen occasionally?
If you "randomly" do a few calls of either
arfima(c(45.4, 3.25, 6.5, 2.14999999999998))
or
arfima(LYGH[[201]])
do you see "random" behavior? [if yes; this confirms typically
memory protection/allocation/... bugs]
fracdiff itself has also been updated (to 1.5-2) but that was really only
a compiler warning, not a bug fix, and I cannot imagine that
that makes the difference.
I think you should consider let your 'Windows Server' provider
update R on that server. Notably the Windows version of R has
had problems fixed in the mean time ..
and at least there's a hope the problem disappears.
Lastly (but probably not helping more), you could use
dput() with control="digits" and even
.Internal(inspect( * )) magic on your LYGH[[201]]
When I use simple numeric vectors "like" your LYGH[..], I get
> dput(c(45.4, 3.25, 6.5, 2.15), control="digits")
c(45.399999999999999, 3.25, 6.5, 2.1499999999999999)
> .Internal(inspect(c(45.4, 3.25, 6.5, 2.15)) + )
@9874728 14 REALSXP g0c3 [] (len=4, tl=0) 45.4,3.25,6.5,2.15
>
Martin
> ________________________________
> From: Martin Maechler <maechler using stat.math.ethz.ch>
> Sent: Thursday, June 1, 2023 1:28 PM
> To: akshay kulkarni <akshay_e4 using hotmail.com>
> Cc: R help Mailing list <r-help using r-project.org>
> Subject: Re: [R] error in arfima...
> >>>>> akshay kulkarni
> >>>>> on Wed, 31 May 2023 20:55:33 +0000 writes:
> > dear members,
> > I am using arfima() from forecast package to model a time
> > series. The following is the code:
> >> LYGH[[202]]
> > [1] 45.40 3.25 6.50 2.15
> >> arfima(LYGH[[202]])
> > Error in .fdcov(x, fdf$d, h, nar = nar, nma = nma, hess = hess, fdf.work = fdf$w) :
> > NA/NaN/Inf in foreign function call (arg 5)
> > I tried viewing .fdcov() with the following code:
> > View(environment(fracdiff)$.fdcov)
> > but I see no stop() with the above mentioned error message.
> > Can you please help?
> Well, as I am the maintainer of the *fracdiff* package,
> I'm trying. OTOH, as we will see below, you did not give us enough
> information to really help you...
> > THanking you,
> > Yours sincerely,
> > AKSHAY M KULKARNI
> Your problem is not (yet) reproducible (*)
> I've done the following in an "empty" (i.e. freshly started) R
> session, R 4.3.0:
> ----begin_R_transcript___save_as__arfima-ex.Rout__------------------------------
> >
> > lN <- loadedNamespaces()
> > library(forecast)
> Registered S3 method overwritten by 'quantmod':
> method from
> as.zoo.data.frame zoo
> > options(width=75)
> > setdiff(loadedNamespaces(), lN) # -- see all the *!^@...! forecast loads:
> [1] "gtable" "dplyr" "tidyselect" "Rcpp" "parallel"
> [6] "scales" "lattice" "ggplot2" "R6" "generics"
> [11] "curl" "lmtest" "tibble" "munsell" "nnet"
> [16] "forecast" "timeDate" "pillar" "rlang" "quantmod"
> [21] "utf8" "urca" "quadprog" "cli" "magrittr"
> [26] "xts" "grid" "nlme" "lifecycle" "fracdiff"
> [31] "vctrs" "glue" "tseries" "zoo" "fansi"
> [36] "colorspace" "TTR" "pkgconfig"
> > arfima(c(45.40, 3.25, 6.50, 2.15))
> Call:
> arfima(y = c(45.4, 3.25, 6.5, 2.15))
> Coefficients:
> d
> 4.583013e-05
> sigma[eps] = 18.01252
> a list with components:
> [1] "log.likelihood" "n" "msg"
> [4] "d" "ar" "ma"
> [7] "covariance.dpq" "fnormMin" "sigma"
> [10] "stderror.dpq" "correlation.dpq" "h"
> [13] "d.tol" "M" "hessian.dpq"
> [16] "length.w" "residuals" "fitted"
> [19] "call" "x" "series"
> >
> > sessionInfo()
> R version 4.3.0 (2023-04-21)
> Platform: x86_64-pc-linux-gnu (64-bit)
> Running under: Fedora Linux 36 (Thirty Six)
> Matrix products: default
> BLAS: /usr/local64.sfs/app/R/R-4.3.0-inst/lib/libRblas.so
> LAPACK: /usr/lib64/liblapack.so.3.10.1
> locale:
> [1] LC_CTYPE=de_CH.UTF-8 LC_NUMERIC=C
> [3] LC_TIME=en_US.UTF-8 LC_COLLATE=de_CH.UTF-8
> [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=de_CH.UTF-8
> [7] LC_PAPER=de_CH.UTF-8 LC_NAME=C
> [9] LC_ADDRESS=C LC_TELEPHONE=C
> [11] LC_MEASUREMENT=de_CH.UTF-8 LC_IDENTIFICATION=C
> time zone: Europe/Zurich
> tzcode source: system (glibc)
> attached base packages:
> [1] graphics grDevices datasets stats utils methods base
> other attached packages:
> [1] forecast_8.21 fortunes_1.5-4 sfsmisc_1.1-15
> loaded via a namespace (and not attached):
> [1] gtable_0.3.3 dplyr_1.1.2 compiler_4.3.0
> [4] tidyselect_1.2.0 Rcpp_1.0.10 parallel_4.3.0
> [7] scales_1.2.1 lattice_0.21-8 ggplot2_3.4.2
> [10] R6_2.5.1 generics_0.1.3 curl_5.0.0
> [13] lmtest_0.9-40 tibble_3.2.1 munsell_0.5.0
> [16] nnet_7.3-19 timeDate_4022.108 pillar_1.9.0
> [19] rlang_1.1.1 quantmod_0.4.22 utf8_1.2.3
> [22] urca_1.3-3 quadprog_1.5-8 cli_3.6.1
> [25] magrittr_2.0.3 xts_0.13.1 grid_4.3.0
> [28] nlme_3.1-162 lifecycle_1.0.3 fracdiff_1.5-2
> [31] vctrs_0.6.2 glue_1.6.2 tseries_0.10-54
> [34] zoo_1.8-12 fansi_1.0.4 colorspace_2.1-0
> [37] TTR_0.24.3 tools_4.3.0 pkgconfig_2.0.3
> >
> ##---end__R_transcript---------------------------------------------------------------
> Note that your error message pointed me to my (old, but still
> fine) package {fracdiff} with its principal function fracdiff()
> around which arfima() is a user-friendly and
> generalization wrapper.
> In other words arfima() calls fracdiff::fracdiff() and the error
> happens there --- for you, but not for me, if I try to use the
> same data as you.
> I see that you must have found that too, because you mentioned
> View(environment(fracdiff)$.fdcov)
> Maybe you need to
> update.packages()
> {which should re-install R packages you have installed but which
> have become outdated in the mean time}
> If the error persists, please send us the output of
> 1)
> dput(LYGH[[202]])
> so we get your exact data
> 2)
> sessionInfo()
> so we get much info about your R "setup"
> .... last but not least: If you are really calling arfima()
> with a time series of length 4 (which your LYGH[[202]] above
> seems to suggest)
> then you really need to learn a bit about the tools you are
> using: It really does not make any sense to fit a somewhat
> sophisticated time-series model (or *any* time-series model, I'd say)
> to a "series" with 4 values.
> Best regards,
> Martin
> --
> Martin Maechler
> ETH Zurich and R Core team
More information about the R-help
mailing list