davidtedfordholt / fable.bsts Goto Github PK

Potential seasonal or date format issue

I think this may have to do with the seasonal special, but we're getting NULL models on yearmonth data when normal dates work. We're also seeing both return NULL models for a seasonal call of 1 month regardless of date format.

library(tidyverse)
library(fabletools)
library(fable)
devtools::install_github('bradisbrad/fable.bsts')
#> Skipping install of 'fable.bsts' from a github remote, the SHA1 (9b263921) has not changed since last install.
#>   Use `force = TRUE` to force installation

source('~/proj/pers/repo/fable.bsts/data-raw/lax_passengers.R')
#> 
#> Attaching package: 'lubridate'
#> The following object is masked from 'package:tsibble':
#> 
#>     interval
#> The following objects are masked from 'package:base':
#> 
#>     date, intersect, setdiff, union
#> `summarise()` regrouping output by 'date' (override with `.groups` argument)

data <- lax_passengers %>% 
  mutate(date = yearmonth(date))

data %>% 
  fabletools::model(
    bsts = fable.bsts::BSTS(passengers ~ intercept() + ar() + seasonal('1 week'))
  )
#> Warning: 2 errors (1 unique) encountered for bsts
#> [2] order.by requires an appropriate time-based object
#> # A mable: 2 x 2
#> # Key:     type [2]
#>   type                  bsts
#>   <chr>              <model>
#> 1 Domestic      <NULL model>
#> 2 International <NULL model>

data <- lax_passengers 

data %>% 
  fabletools::model(
    bsts = fable.bsts::BSTS(passengers ~ intercept() + ar() + seasonal('1 week'))
  )
#> # A mable: 2 x 2
#> # Key:     type [2]
#>   type             bsts
#>   <chr>         <model>
#> 1 Domestic       <BSTS>
#> 2 International  <BSTS>

data %>% 
  fabletools::model(
    bsts = fable.bsts::BSTS(passengers ~ intercept() + ar() + seasonal('1 month'))
  )
#> Warning: 2 errors (1 unique) encountered for bsts
#> [2] nseasons == round(nseasons) is not TRUE
#> # A mable: 2 x 2
#> # Key:     type [2]
#>   type                  bsts
#>   <chr>              <model>
#> 1 Domestic      <NULL model>
#> 2 International <NULL model>

^{Created on 2020-09-09 by the reprex package (v0.3.0)}

We will need to figure out how to specify a prior intuitively. This includes the question of how much would need to be specified BEFORE the fable workflow, and whether or not things need to be in a tibble with the same key structure, a tsibble with the same key structure and index values, maybe just partial index values, or as columns in the tsibble, itself... I don't know!

This will be alongside SdPrior and all the other specific prior specification methods.

From Steven Scott:

There should only be (at most) one non-stationary trend model and one stationary trend model. Multiple seasonal models are okay, as long as the seasonal periods are different. The shorter and more "boring" your data, the fewer state models you should have. E.g. you could have a local linear trend plus an AR, but in the airline data (for example) that's probably asking for trouble.

Honestly, I'm not sure if this is applicable, but it is an additive model

I'm not entirely sure that we will be able to warn for all of them, but if we find model specifications that don't make sense or don't run we need to document them and create warnings for them. It would be nice if those warnings just returned an empty mbl object instead of erroring the process out.

It seems like I should use this. And probably other tools for checking package builds and coverage and whatnot.

I'm getting the following error when trying to install the package:

> remotes::install_github('https://github.com/davidtedfordholt/fable.bsts')
Downloading GitHub repo davidtedfordholt/fable.bsts@HEAD
✓  checking for file ‘/private/var/folders/wl/klf5ry4n2vg4lfmtqk5tv5sr0000gn/T/RtmppBnAzW/remotesaf2e146d3d0e/davidtedfordholt-fable.bsts-4488a12/DESCRIPTION’ ...
─  preparing ‘fable.bsts’:
✓  checking DESCRIPTION meta-information ...
─  checking for LF line-endings in source and make files and shell scripts
─  checking for empty or unneeded directories
─  building ‘fable.bsts_0.1.0.tar.gz’
   
* installing *source* package ‘fable.bsts’ ...
** using non-staged installation via StagedInstall field
** R
** byte-compile and prepare package for lazy loading
Error in is_regular_interval(x) && sum(vec_c(!!!vec_data(x))) == 0 : 
  invalid 'x' type in 'x && y'
Error: unable to load R code in package ‘fable.bsts’
Execution halted
ERROR: lazy loading failed for package ‘fable.bsts’

In the following example, the dtwclust library is explicitly loaded into each worker. I'm not sure how to explicitly load things like rlang into the workers when using the implementation of the future packages in fable. @mitchelloharawild , I assume I'm just failing to put the required packages someplace appropriate, to ensure that they get called?

require("doParallel")
# Create parallel workers
workers <- makeCluster(2L)
# Preload dtwclust in each worker; not necessary but useful invisible(clusterEvalQ(workers, library("dtwclust")))
# Register the backend; this step MUST be done registerDoParallel(workers)

Package installation of fable.bsts fails with the message:

Error : package or namespace load failed for ‘future’:
 .onLoad failed in loadNamespace() for 'globals', details:
  call: asNamespace(ns, base.OK = FALSE)
  error: not a namespace
Error: unable to load R code in package ‘fable.bsts’
Execution halted

The package future exists on my system and is up to date. It was successfully installed via install.packages.

My system:

• R version: 4.0.0
• Platform: macOS 10.15.4 (Catalina)

Please briefly describe your problem and what output you expect. If you have a question, please don't use this form. Instead, ask on https://stackoverflow.com/.

Please include a minimal reproducible example (AKA a reprex). If you've never heard of a reprex before, start by reading https://www.tidyverse.org/help/#reprex.

I encountered the following error while installing fable.bsts on my AWS SageMaker notebook instances. Can you please help me with this? Thanks!

Below is the R code for installation

ipak <- function(pkg){
  new.pkg <- pkg[!(pkg %in% installed.packages()[, "Package"])]
  if (length(new.pkg))
    install.packages(new.pkg, dependencies = TRUE)
  try(sapply(pkg, require, character.only = TRUE), silent = TRUE)
}
packages <- c("fable", "tsibble", 
              "tsibbledata", 
              "lubridate", "dplyr", 
              "ggplot2", "stringi","fable.prophet", "remotes")
ipak(packages)

remotes::install_github("davidtedfordholt/fable.bsts")

Please see below for the error message

_Downloading GitHub repo davidtedfordholt/fable.bsts@HEAD

Boom (NA -> 0.9.11) [CRAN]
timechange (NA -> 0.1.1 ) [CRAN]
lubridate (1.8.0 -> 1.9.0 ) [CRAN]
BoomSpike... (NA -> 1.2.5 ) [CRAN]
bsts (NA -> 0.9.9 ) [CRAN]
Installing 5 packages: Boom, timechange, lubridate, BoomSpikeSlab, bsts

Updating HTML index of packages in '.Library'

Making 'packages.html' ...
done

Running R CMD build...

• checking for file ‘/tmp/RtmpQwSs0m/remotes27a876a9353a/davidtedfordholt-fable.bsts-4488a12/DESCRIPTION’ ... OK
• preparing ‘fable.bsts’:
• checking DESCRIPTION meta-information ... OK
• checking for LF line-endings in source and make files and shell scripts
• checking for empty or unneeded directories
  Omitted ‘LazyData’ from DESCRIPTION
• building ‘fable.bsts_0.1.0.tar.gz’
  Warning message in i.p(...):
  “installation of package ‘/tmp/RtmpQwSs0m/file27a8520733d4/fable.bsts_0.1.0.tar.gz’ had non-zero exit status”_

The environment info is here

R version 4.2.2 (2022-10-31)
Platform: x86_64-conda-linux-gnu (64-bit)
Running under: Amazon Linux 2

Matrix products: default
BLAS/LAPACK: /home/ec2-user/anaconda3/envs/R/lib/libopenblasp-r0.3.21.so

locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
[3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
[5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=en_US.UTF-8 LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats graphics grDevices utils datasets methods base

other attached packages:
[1] remotes_2.4.2 fable.prophet_0.1.0 Rcpp_1.0.9
[4] bsts_0.9.9 xts_0.12.2 zoo_1.8-11
[7] BoomSpikeSlab_1.2.5 Boom_0.9.11 stringi_1.7.8
[10] ggplot2_3.4.0 dplyr_1.0.10 lubridate_1.9.0
[13] tsibbledata_0.4.1 tsibble_1.1.3 fable_0.3.2
[16] fabletools_0.3.2

loaded via a namespace (and not attached):
[1] pbdZMQ_0.3-8 tidyselect_1.2.0 repr_1.1.4
[4] purrr_0.3.5 lattice_0.20-45 colorspace_2.0-3
[7] vctrs_0.5.1 generics_0.1.3 htmltools_0.5.3
[10] base64enc_0.1-3 utf8_1.2.2 rlang_1.0.6
[13] pillar_1.8.1 glue_1.6.2 withr_2.5.0
[16] DBI_1.1.3 rappdirs_0.3.3 distributional_0.3.1
[19] uuid_1.1-0 lifecycle_1.0.3 munsell_0.5.0
[22] anytime_0.3.9 gtable_0.3.1 evaluate_0.18
[25] fastmap_1.1.0 curl_4.3.3 fansi_1.0.3
[28] IRdisplay_1.1 scales_1.2.1 IRkernel_1.3
[31] jsonlite_1.8.3 farver_2.1.1 digest_0.6.30
[34] grid_4.2.2 cli_3.4.1 tools_4.2.2
[37] magrittr_2.0.3 tibble_3.1.8 crayon_1.5.2
[40] tidyr_1.2.1 pkgconfig_2.0.3 ellipsis_0.3.2
[43] MASS_7.3-58.1 assertthat_0.2.1 R6_2.5.1
[46] compiler_4.2.2

Because people like to view things!