stat_ellipse Rdocumentation on vcov

[Abiologist]

Dear All - I was surprised to find that the function ggplot2::stat_ellipse uses a different covariate (vcov) function for the t-distribution setting (robust) and the normal distribution setting (non-robust).
While I'm sure this is correct - there is no mention of this in the Rdocumentation - which must result in quite a few pleasant surprises when the t ellipses are smaller than the normal ellipses !

Perhaps this could be mentioned (?) - also perhaps an argument for vcov to allow changing this ? (plus default behaviour as above if vcov not specified).

[teunbrand]

Thanks for the report! Just to clarify, the request here is to put a note in the documentation that type = "t" will use MASS::cov.trob(), whereas others use stats::cov.wt()?

[Abiologist]

yes - with perhaps a note saying that the robust vcov sometimes results in smaller ellipses ! Additionally- perhaps another argument could be added to force change in this behaviour (if you think this appropriate) - many thanks

…

On Mon, 19 May 2025 at 11:39, Teun van den Brand ***@***.***> wrote: *teunbrand* left a comment (tidyverse/ggplot2#6462) <#6462 (comment)> Thanks for the report! Just to clarify, the request here is to put a note in the documentation that type = "t" will use MASS::cov.trob(), whereas others use stats::cov.wt()? — Reply to this email directly, view it on GitHub <#6462 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADYLLF2H3GCW4P27352TEI327GRELAVCNFSM6AAAAAB5NGQTS2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQOJQGM2DMOBQGA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[teunbrand]

I don't think stat_ellipse() is the right place to explain the various nuances of the choices for the same reason I don't think stat_density() is the right place to discuss the various nuances of different bandwidth selectors. If we link the relevant functions, people can read up and inform their choices.

[Abiologist]

yes - teunbrand's note plus one or a couple of links sounds good.