Currently the boxes change color when F exceeds 1:
GetMSbetweenColor <- function(owp) {
if (owp$stats$F.statistic > 1) {
return(brewer.pal(n = 8, name = "Paired")[5])
}
else {
return(brewer.pal(n = 8, name = "Paired")[2])
}
}
GetMSwithinColor <- function(owp) {
if (owp$stats$F.statistic > 1) {
return(brewer.pal(n = 8, name = "Paired")[6])
}
else {
return(brewer.pal(n = 8, name = "Paired")[1])
}
}
If the thinking is that the box colors should "alert" the reader, it makes more sense to me to alert the reader to the presence of a significant F-statistic. So, the color change behavior should depend on
if (owp$stats$F.statistic > F.critical)
where we likely grab F.critical from summary model output info and a call to an F-distribution probability function.
In my own use of the granovagg.1w for exploratory data analysis, I've found that the visual property of the boxes does little to give me a sense of how close I am to the threshold of a significant effect. If anything, having a reliable color-change at the significance threshold (instead of the > 1 threshold) would help.
Currently the boxes change color when F exceeds 1:
If the thinking is that the box colors should "alert" the reader, it makes more sense to me to alert the reader to the presence of a significant F-statistic. So, the color change behavior should depend on
where we likely grab
F.critical
from summary model output info and a call to an F-distribution probability function.In my own use of the
granovagg.1w
for exploratory data analysis, I've found that the visual property of the boxes does little to give me a sense of how close I am to the threshold of a significant effect. If anything, having a reliable color-change at the significance threshold (instead of the > 1 threshold) would help.@rmpruzek and @wildoane, what are your thoughts?