Transpose coords

[xrobin]

Coords returns a matrix with thresholds in columns, and the measurement in rows.

This has always been a bit weird but is becoming problematic with pipelines where a data.frame in transposed form would be better suited.

Expected behavior

> library(dplyr)
> roc(aSAH, outcome, wfns) %>% coords()
Setting levels: control = Good, case = Poor
Setting direction: controls < cases
                threshold specificity sensitivity
      -Inf   0.0000000   1.0000000
       1.5   0.5138889   0.9512195
       2.5   0.7916667   0.6585366
       3.5   0.8333333   0.6341463
       4.5   0.9444444   0.4390244
       Inf   1.0000000   0.0000000

[xrobin]

See wiki page Updating coords for an assessment of potential impact on dependent packages.

[xrobin]

• TODO: add an FAQ entry documenting this change, the warning, and how to silence it.

[xrobin]

Email sent to 49 maintainers of reverse dependencies.

Delivery failures for EFS (Ursula Neumann, mailbox quota exceeded), blkbox (Boris Guennewig, 550 Address rejected), and sambia (Norbert Krautenbacher, 550 Recipient address rejected).

[xrobin]

4 emails and 2 PR sent to authors of packages found using coords:

• Use tidy version of coords (xrobin/pROC#54) tidymodels/yardstick#110
• Use tidy version of coords (xrobin/pROC#54) waldronlab/doppelgangR#19

[xrobin]

Email sent now to 57 reverse depends/imports (5 delivery failures) and 21 reverse suggests (1 delivery failure).

Reminder diff sent to Colin Tredoux (r4lineups) and bug report for biomod2 (bug #6650), which are the last 2 packages detected as impacted that haven't been fixed yet.

[xrobin]

The coords function now uses a data.frame internally too. Following Tibble's Overview, I dropped row names which are sometimes duplicated, and turned off dropping dimensions by default. This required some adaptation of the tests and examples.

[xrobin]

The use of a data.frame internally came with a significant performance hit (apparently related to row indexing/updating operations, which are used a lot here). CI operations became unacceptably slow. Reverted to use matrix internally and added an as.matrix option, used in CI. Still defaults to returning a data.frame, but conversion happens in the end.

[xrobin]

pROC 1.16.0 is on CRAN with transpose = FALSE by default.

Closing, future changes to remove the warning (probably in ~ 1 year) and potentially drop transpose = TRUE altogether will go to new issues.