Figure 1:

Rashomon sets and model class reliance – Panel (A) illustrates a hypothetical Rashomon set $\mathcal{R}\left(ϵ\right)$, within a model class $\mathcal{F}$. The y-axis shows the expected loss of each model $f\in \mathcal{F}$, and the x-axis shows how much each model f relies on X₁ (defined formally in Section 3). Along the x-axis, the population-level MCR range is highlighted in blue, showing the values of MR corresponding to well-performing models (see Section 4). Panel (B) shows the in-sample analogue of Panel (A). Here, the y-axis denotes the in-sample loss, $\widehat{\mathbb{E}}L(f,Z):=\frac{1}{n}{\sum }_{i=1}^{n}L(f,{\mathbf{Z}}_{[i,\cdot ]})$; the x-axis shows the empirical model reliance of each model $f\in \mathcal{F}$ on X₁ (see Section 3); and the highlighted portion of the x-axis shows empirical MCR (see Section 4).