glance on an aov model reports the wrong df value

[cmartin]

Unless I'm misunderstanding something, glance should report the same df as in summary (e.g. 2 in this example) but it seems to always report df+1

library(broom)

d <- data.frame(
  y = c(3,4,3.5,4.1,6,7),
  group = c("a","b","a","b","c","c")
)
m <- aov(y~group,data = d)

summary(m)
glance(m)

This is with broom 0.4.2 (the current version with tidyverse if I'm not mistaken)

[kcmacstat]

Same problem occurs with glance on an lm model, i.e., glance(lm.object). The df is computed incorrectly even though summary(lm.object) shows the correct df.

[lemna]

glance() should probably return the numerator and denumerator df for the F statistic. It doesn't do that because it uses the df slot from summary.lm() (which contains the rank of the model matrix). It might be more sensible to use the df from the fstatistic slot.

[alexpghayes]

In #624 I've moved to the numerator degrees of freedom for the F-statistic for lm objects. Here, I'm a little less certain about what to do. In a particular ANOVA table there are potentially many F-tests of interest, all with different degrees of freedom. So far the approach has been to treat the aov like the corresponding regression, and report the statistics for the regression, but I'm not sure how I feel about this. I almost think glance.aov() should be an error instead. Thoughts @dgrtwo?

[alexpghayes]

Tentatively I've thrown an error.

[simonpcouch]

Addressed in 0.7.0. :-)

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.