In using ggtern, I think the region plotted by geom_confidence_tern may be what is more normally termed a prediction region, rather than a confidence region? i.e. probabilities relate to a new data point, rather than to a population parameter?
My reasoning is analogous to this discussion for stat_ellipse in ggplot2. In particular, they don't get tighter with increasing sample size, and seem to roughly contain the proportion of the data points indicated by the probability values.
In using
ggtern, I think the region plotted bygeom_confidence_ternmay be what is more normally termed a prediction region, rather than a confidence region? i.e. probabilities relate to a new data point, rather than to a population parameter?My reasoning is analogous to this discussion for
stat_ellipseinggplot2. In particular, they don't get tighter with increasing sample size, and seem to roughly contain the proportion of the data points indicated by the probability values.References https://en.wikipedia.org/wiki/Prediction_interval https://github.com/tidyverse/ggplot2/issues/2776