hBayesDM package error in R 4.2.2

[AaronLaycock]

Hi all,

FYI

After experiencing difficulties (PSA) running an ORL model using the hBayesDM package in R version 4.2.2 (with the appropriate version of Rtools), the issue was rectified when I regressed to R version 4.1.3 (and associated Rtools package).

[plonskipe]

having trouble with dd_hyperbolic too. I think it's related to https://stackoverflow.com/questions/72848442/r-warning-lengthx-2-1-in-coercion-to-logical1
"example" data runs fine but input data frames result in the error:
Error in is.null(data) || is.na(data) :
'length = 8568' in coercion to 'logical(1)'
where the length is the number of cells in the data frame. Clash with a single logical on the other side.

[steffen-stell]

The problem is due to this line:

hBayesDM/R/R/hBayesDM_model.R

Line 144 in 534907d

if (is.null(data) || is.na(data) || data == "") {

This test makes no sense. Both is.na(data) and data == "" produce matrices of element-wise checks for a data frame input. In older versions of R these were coerced to scalar by just taking the first element of the matrix. In newer versions it just results in an error, because if(), || and && take only scalars. If you want to check for scalar NA, this should be something like

    if (is.null(data) || identical(data, NA) || identical(data, "")) {

or if there must be no NAs at all in the data frame then

    if (is.null(data) || any(is.na(data)) || identical(data, "")) {

[youngahn]

Thanks @steffen-stell for your suggestion. We will double check and fix the line in the next version.

[ChristinaWUUK]

Hi everyone, I encounter the same issue. is there a version that is now working or do you know how to fix this? many thanks

[AaronLaycock]

Hey ChristinaWUUK, I am not aware of when to expect any updates to the package. That said, I did have to re-run some analysis recently and the package worked fine once I reverted back to the appropriate archived versions of R and Rtools.

[youngahn]

Hi @ChristinaWUUK, could you check if this would work?

devtools::install_github("CCS-Lab/hBayesDM", ref="develop", subdir="R")

[lukasgunschera]

@youngahn I am experiencing the same issue - for me, using the above installation does not resolve the

Error in is.null(data) || is.na(data) : 'length = 16000' in coercion to 'logical(1)'

[ChristinaWUUK]

Thank you @youngahn, really appreciate your promt response to help! I tried installing from github, however, I later on got stuck with the running bandit2arm_delta (Error in stanc(file = file, model_code = model_code, model_name = model_name, :
Empty or invalid filename!). I ended up installing a previous version of R (4.1.3) and this worked @lukasgunschera

[ChristinaWUUK]

@AaronLaycock thanks! which version of R and R tools did you use? It worked fine with R 4.1.3 and Rtools 4.0 but i did have to install install.packages("StanHeaders", repos = c("https://mc-stan.org/r-packages/", getOption("repos")))
install.packages("rstan", repos = c("https://mc-stan.org/r-packages/", getOption("repos")))

[AaronLaycock]

@ChristinaWUUK, no worries. Great news, it's working. I used R version 4.1.3 with Rtools 4.0 also, it ran with no problems from the initial download.

[lukasgunschera]

Thank you @ChristinaWUUK @AaronLaycock - it works fine after reverting to R 4.1.3.
Has anyone attempted to implement @steffen-stell's solution in 4.3.3?

[steffen-stell]

Since people keep asking and this is a fairly simple issue to fix, I have made PR #161. You can install it like this:

devtools::install_github(
  repo = "CCS-Lab/hBayesDM",
  ref = "627c1ea35121b85760fdddec68583e82431c06e6", 
  subdir = "R"
  )