Fix failing checks

[sjentsch]

These changes should take care of the failing checks on GitHub (oldrel and devel). Ironically, they do only partially (see below why).

To prevent the checks for devel from failing, the Durbin(-Conover)-test was re-implemented based upon NIST and the Wikipedia entry for the test. This removes the dependency on PMCMR which is deprecated (and will be removed with the next version of R). devtools::test() for anovarmnp is successful. I conducted further tests with examples from the internet, comparing it with the current version of jamovi (using PMCMR) and the re-implementation, ensuring the same results.

The GitHub actions for macos-latest and ubuntu-oldrel fail (because of install.packages issue). I can't check the macos-issue, however, when trying the code out in a docker / rocker container with R 4.4 (~ ubuntu-oldrel) it works without problems.

[jonathon-love]

The GitHub actions for macos-latest and ubuntu-oldrel fail

i think we'll probably just standardise on windows, and only bother running the windows tests ... the macOS and linux tests often fail for reasons outside of our control ...

[raviselker]

A couple of suggestions and questions, but looks good!

[raviselker]

The GitHub actgions for the ubuntu-latest (devel) fails on the actual tests. Any idea why this is happening?

[sjentsch]

Regd.: The GitHub actgions for the ubuntu-latest (devel) fails on the actual tests. Any idea why this is happening?

No, not yet. I also just observed it. Strangely enough, the error did not happen in my branch (which was branched from an older version of jmv; see screenshot below).
It is unlikely that it has to do with a change to R-devel happened in that period between my submission and the PR (both on Jan 16). It is more likely that it has to do with code you added recently (fix_failingchecks is 15 commits behind main).

[shun2wang]

I think I can fix failure on macOS, but I have no access to push on your PR so I made another PR see:#444

[shun2wang]

Failed on Linux older version it is becasue RcppEigen package build error. I guess it won't compile against R 4.4.x.

[jonathon-love]

yeah, i think testing across all these OSes doesn't actually buy us anything ... it doesn't increase the likelihood that something problematic in our code will be identified ... all it does is raise the maintenance burden.

The GitHub actgions for the ubuntu-latest (devel) fails on the actual tests. Any idea why this is happening?

i think i'm in the "who knows? who cares?" camp ... it's almost definitely because of something in R which needs babying, and almost definitely not anything that we need to care about.

[sjentsch]

@raviselker: I resolved the comments / suggestions that you made previously. Please have another look!

[raviselker]

Just one more comment. Rest looks good to me!

[sjentsch]

@raviselker: I dealt with your comment.

[raviselker]

Looks good!

[jonathon-love]

OK, i've squashed and merged this, but i did leave a small change to descriptives behind ...

If this is a fix for something, let me know and i'll merge it as a separate commit.

# before
counts <- do.call("[", c(list(freq), grid[row, ]))

# after
counts <- as.numeric(do.call("[", c(list(freq), grid[row, ])))

[sjentsch]

@jonathon-love: The as.numeric-line you removed makes that the unit tests under R 4.6 fail.

[sjentsch]

Error (testdescriptives.R:211:5): Descriptives works old scenario
Error in `as.character.factor(X, ...)`: malformed factor
Backtrace:
    ▆
 1. ├─desc$frequencies[[1]]$asDF at testdescriptives.R:211:5
 2. ├─jmvcore:::`$.ResultsElement`(desc$frequencies[[1]], asDF) at testdescriptives.R:211:5
 3. └─jmvcore (local) `<fn>`()
 4.   └─jmvcore:::as.data.frame.Table(self)
 5.     └─base::unlist(as.list(column), use.names = FALSE)
 6.       └─base (local) URapply(x, as.character)
 7.         └─base::rapply(x, Fn, how = "list")
 8.           ├─base (local) `<fn>`(X, ...)
 9.           └─base::as.character.factor(X, ...)

[sjentsch]

Without as.numeric:

> descriptives(data, vars=c("w", "y", "z"), splitBy = "x",
                              freq=TRUE, median=TRUE, mode=TRUE, skew=TRUE,
                              kurt=TRUE, pc=TRUE)$frequencies[[1]]

 Frequencies of w                                    
 ─────────────────────────────────────────────────── 
   w    x    Counts     % of Total    Cumulative %   
 ─────────────────────────────────────────────────── 
   1    a    [table]       0.00000         0.00000   
        b    [table]       0.00000         0.00000   
        c    [table]       0.00000         0.00000   
   2    a    [table]       0.00000         0.00000   
        b    [table]       0.00000         0.00000   
        c    [table]       0.00000         0.00000   
   3    a    [table]       0.00000         0.00000   
        b    [table]       0.00000         0.00000   
        c    [table]       0.00000         0.00000   
 ─────────────────────────────────────────────────── 

There were 18 warnings (use warnings() to see them)

[sjentsch]

With as.numeric:

> descriptives(data, vars=c("w", "y", "z"), splitBy = "x",
                              freq=TRUE, median=TRUE, mode=TRUE, skew=TRUE,
                              kurt=TRUE, pc=TRUE)$frequencies[[1]]

 Frequencies of w                                     
 ──────────────────────────────────────────────────── 
   w    x    Counts      % of Total    Cumulative %   
 ──────────────────────────────────────────────────── 
   1    a    2.000000      16.66667        16.66667   
        b    1.000000       8.33333        25.00000   
        c    1.000000       8.33333        33.33333   
   2    a    1.000000       8.33333        41.66667   
        b    2.000000      16.66667        58.33333   
        c    1.000000       8.33333        66.66667   
   3    a    1.000000       8.33333        75.00000   
        b    1.000000       8.33333        83.33333   
        c    2.000000      16.66667       100.00000   
 ──────────────────────────────────────────────────── 

[sjentsch]

NB: The as.numeric works in 4.4 and 4.5 too.

[jonathon-love]

roger that. i've brought that across now.