Error when doing autocorrelation diagnostic about shinystan

[jgabry]

Hmm, is this something that's only happening for this particular model or does it happen when you use shinyStan with any model?

If your stanfit object isn't too big then you could share a dropbox link or google drive or something (even github). If not then the best option is probably to make a post to the Stan users forum because it will let you upload your model code, data, etc.

[dalejordan]

Ok, I uploaded the saved fit object to dropbox:
https://www.dropbox.com/s/c837lzhws5s5y9v/auto-error.gz?dl=0

This is the first model I tried to use shinystan with.

Let me know if you need any of the model or data files.

[jgabry]

Thanks for sending. I'm actually not getting those errors with your stanfit object.

A few follow up questions:

• Do the multiview and autocorrelation pages cause this error if you use the shinyStan demo instead of your own model? (you can run that with launch_shinystan_demo() and then enter 1 to use the default demo).
• Do have R version 3.1.1 or 3.1.2?

My first thought is that something went wrong during installation. Can you try reinstalling shinyStan? Except instead of installing it the way it suggests in the installation instructions try installing it straight from the 'develop' branch:

# restart R session
remove.packages("shinyStan")
devtools::install_github("stan-dev/shinystan", ref = "develop")
# restart R session
library(shinyStan)

Let me know if that fixes the issue and if not I'll come up with a few other things to try.

[dalejordan]

I re-installed shinyStan from github as you suggested. The demo failed with the same diagnostics as before as did my model.

I am running R ver. 3.1.2

[dalejordan]

I also got the following warnings after the first launch; I don't know if they are relevant.
Loading...
For large models shinyStan may take a few moments to launch.

Listening on http://127.0.0.1:6921
Loading required package: zoo

Attaching package: ‘zoo’

The following objects are masked from ‘package:base’:

as.Date, as.Date.numeric

Attaching package: ‘xts’

The following objects are masked from ‘package:dplyr’:

first, last

[jgabry]

First of all, sorry it's not working for you! If you're willing to keep trying a few things we'll get this sorted out.

Regarding those warnings:

• The first one about large models is just a message letting people know it might take a bit to start up for really big models.
• The ones about objects being masked from packags are letting you know that packages that are loaded have functions with the same names as other functions from other loaded packages. This can be problematic in some circumstances but shouldn't be here. (Just in case you can try not loading dplyr before running shinyStan, but I've never had a problem with running shinyStan with dplyr loaded.)

The next thing that would be helpful is if you can doing the following:

1. launch shinyStan
2. get it to fail again on autocorrelation or multiview
3. without restarting R run sessionInfo() and copy the output that gives you and send it?

sessionInfo() should just give you a bunch of text listing packages and other info. And you'll be able to just copy and paste the text. It's possible I'll see something in sessionInfo() that might be problematic, but if not it will at least allow me to rule out a few possibilities.

A few other comments:

• I still don't think it's a linux issue, but I will double-check with some of the other Stan developers who are more familiar with linux than I am.
• If I can't figure out what's going soon then while I'm working on it I'll make a branch on GitHub that has all of shinyStan except for the parts giving you errors so you can install from that branch and at least have a version that doesn't fail on you.

[dalejordan]

Here's the output after the demo fails:

sessionInfo()
R version 3.1.2 (2014-10-31)
Platform: x86_64-unknown-linux-gnu (64-bit)

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] parallel stats graphics grDevices utils datasets methods
[8] base

other attached packages:
[1] xtable_1.7-4 threejs_0.2.1 xts_0.9-7 zoo_1.7-11
[5] dygraphs_0.4.1 reshape2_1.4.1 gtools_3.4.1 dplyr_0.4.0
[9] plyr_1.8.1 rstan_2.5.0 inline_0.3.13 Rcpp_0.11.3
[13] shinyStan_1.0.0 shinyBS_0.50 shiny_0.11.1 ggplot2_1.0.0

loaded via a namespace (and not attached):
[1] assertthat_0.1 base64_1.1 colorspace_1.2-4 compiler_3.1.2
[5] DBI_0.3.1 digest_0.6.8 grid_3.1.2 gtable_0.1.2
[9] htmltools_0.2.6 htmlwidgets_0.3.2 httpuv_1.3.2 labeling_0.3
[13] lattice_0.20-29 lazyeval_0.1.10 magrittr_1.5 MASS_7.3-35
[17] mime_0.2 munsell_0.4.2 proto_0.3-10 R6_2.0.1
[21] rjson_0.2.15 RJSONIO_1.3-0 scales_0.2.4 stats4_3.1.2
[25] stringr_0.6.2 tcltk_3.1.2 tools_3.1.2 yaml_2.1.13

[jgabry]

Ok thanks. I'll take a look.

On the advice of Stan developer @bgoodri, can you execute options(error = recover) in R, and after doing that run shinyStan again and trigger the error. R should then give you some output about the error that you can copy and send.

[dalejordan]

Here's the output:

launch_shinystan_demo()

Listening on http://127.0.0.1:6921
Error in eval(substitute(expr), envir, enclos) :
invalid subscript type 'closure'
Error: invalid subscript type 'closure'

Enter a frame number, or 0 to exit

1: launch_shinystan_demo()
2: launch_demo(eight_schools)
3: shiny::runApp(system.file("shinyStan", package = "shinyStan"))
4: shinyCallingHandlers(while (!.globals$stopped) {
serviceApp()
Sys.sle
5: withCallingHandlers(expr, error = function(e) {
handle <- getOption("shi
6: serviceApp()
7: service(timeout)
8: run(timeoutMs)

Selection:

[jgabry]

Ok, I think I found at least part of the problem. I implemented what I hope is the solution on a new github branch. Can you trying installing from that branch by running:

devtools::install_github("stan-dev/shinystan", ref = "feature/autocorr_fix")

Then restart R and give shinyStan a try one more time and see if you still get the errors.

[dalejordan]

Great! All fixed in the new version. Thanks for the great work.

(I noticed I no longer get the the masked dplyr functions warning. Was this related to the problem?)

[jgabry]

Ok, great! Yes, it actually did end up being related to dplyr, or more precisely there was a typo (an omission really) in the code for function that produces the autocorrelation plots that would only cause a problem if dplyr is loaded, and still then it would only cause a problem if the packages were loaded in a particular order. It's a bit surprising actually that you're the first person to get this error (or at least the first to report it). So while I'm sorry it wasn't working for you I'm also glad since I wouldn't have noticed it otherwise. I'll be adding this fix to the next official release.

Thanks for your patience and for bringing this bug to my attention!

[jgabry]

I'm going to go ahead and close this issue and add this bug fix to the main branch for the package.