nice() ANOVA table MSE report

[m-Py]

Hello,

first of all, thanks a lot for creating the afex package. I am using it a lot to compute
all of my ANOVAs or ANCOVAs. I have a question / suggestion with regard to the ANOVA
table that is produced when conducting an ANOVA with afex. When I compute a
multi-factorial ANOVA with several between-subjects factors, I get a table like this:

> # ANOVA example using Iris data

> data(iris)
> iris$id <- 1:nrow(iris)
> iris$rndFactor <- factor(rep(sample(3), 50)) # some second random between-subjects factor
> aoov <-  afex::aov_ez("id", "Sepal.Length", iris, between = c("Species", "rndFactor"))
> nice(aoov)
Anova Table (Type 3 tests)

Response: Sepal.Length
             Effect     df  MSE          F   ges p.value
1           Species 2, 141 0.27 119.83 ***   .63  <.0001
2         rndFactor 2, 141 0.27       0.06 .0008     .95
3 Species:rndFactor 4, 141 0.27       1.47   .04     .21

In this case, the column MSE contains the mean squared error three times. Is this
intended behavior? Maybe it would be more appropriate if the mean squared estimate
for each between-subject factor is reported in this column instead? I am asking because
when using rmarkdown I can directly use the nice() table in a pdf report, and
in this case I do not like to use the MSE column as it is currently implemented. But
maybe I am missing something and this is in fact intended behavior.

Best regards and thanks a lot,
Martin

[singmann]

I am not sure I understand. What would you prefer to see there? Can you perhaps show this in the example (i.e., what kind of values).

[m-Py]

Hello,

I mean the mean square between estimate for each factor. I would get those from the afex object like this:

# continuing the example from above
aoov$Anova[["Sum Sq"]] / aoov$Anova[["Df"]]

I think these values are also reported in JASP or SPSS; however it is of course not necessary to report them in the nice() table by default, and I can produce them for myself if I want to. I just wondered whether it was intentional that MSE is reported in the row for each between-subjects factor (and the mean squared between estimate would be an alternative).

Best regards,
Martin

[singmann]

Hmm, this is probably what I wanted to have there. No clue, why I do not have it there. Let me think about it.