Closed IndrajeetPatil closed 1 month ago
You're right, of course. The dependency bump in estimability is because in R version 4.3, responsibile estimability checking was FINALLY added the predict.lm
. That kind of obviates the need for estimability::epredict()
.
I hadn't thought carefully about the effect of this on emmeans, and I'd be inclined -- if possible -- to try to put through an update of estimability that does not require as high an R version. I have to look carefully to ensure that estimability will actually work with lower R versions, rather than bumping up the requirements for emmeans. I'll let you know soon.
Softening the required R version to 4.1, if possible, would be an even better solution!
Thanks a lot.
The updated estimability 5.1 was just submitted to CRAN. Or you can install it from GitHub.
Thanks for quickly resolving this!
Thanks for calling it to my attention. This put some serious constraints on any other package depending on estimability.
{estimability}
is a runtime or hard dependency of{emmeans}
, which needsR 4.3
.This means that anyone on
R < 4.3
system can't actually install{emmeans}
. But this is not reflected in the minimum needed R version specified inDESCRIPTION
:https://github.com/rvlenth/emmeans/blob/420328de19d9916135ed84b64b7ee5252e2a50ec/DESCRIPTION#L17