Plot LOO Comparison: Scatter and Barplot — plot.uncounted_loo

Produces diagnostic plots comparing observation influence across two models.

Usage

# S3 method for class 'uncounted_loo_compare'
plot(x, type = c("scatter", "bar"), n = 20, label_top = 10, ...)

Arguments

x: An "uncounted_loo_compare" object.
type: "scatter" (default) or "bar".
n: Number of top observations to show in barplot (default 20).
label_top: Number of points to label in scatter (default 10).
...: Additional arguments passed to plot.

Details

Scatter plot (type = "scatter"). Each point is one observation/country. The x-axis is \(\%\Delta\xi\) for Model 1, the y-axis for Model 2. The diagonal line marks equal influence. Points near the origin are non-influential under both models. Points far from the diagonal have divergent influence – they affect the two models differently. The top label_top most influential points are labeled. See compare_loo for quadrant interpretation.

Bar plot (type = "bar"). Horizontal side-by-side bar chart of the top n most influential observations (by maximum absolute percentage change across both models). Red bars are Model 1, blue bars are Model 2. Useful for quickly seeing which observations matter most and whether the models agree on direction.