Make nns_m_reg plot the faithful R residual plot

R's NNS.M.reg plot output is its residual plot (Multivariate_Regression.R:367-377):
actual y over the observation index as steelblue open circles, fitted y.hat as a
red line, with a pink (alpha 0.375) confidence band when present. The previous
fitted-vs-actual scatter / residuals-about-zero scatter did not match R.

Both plot=True and residual_plot=True now render this single faithful figure.
Strengthens the m_reg plot test to assert the steelblue-actual / red-fitted colors.

Author: claude

src/nns/multivariate_regression.py

@@ -116,41 +116,40 @@ def nns_m_reg(
         "Fitted.xy": fitted,
     }
     if plot or residual_plot:
-        _render_m_reg(fitted, plot=plot, residual_plot=residual_plot)
+        _render_m_reg(fitted)
     return result
 
 
-def _render_m_reg(
-    fitted: dict[str, NDArray[np.float64] | NDArray[np.str_]],
-    *,
-    plot: bool,
-    residual_plot: bool,
-) -> None:
-    """Render multivariate fitted-vs-actual / residual diagnostics on ``plot=True``.
-
-    The synthetic predictors make a single x-axis ill-defined, so ``plot`` shows
-    fitted vs actual (steelblue points, red 1:1 line) and ``residual_plot`` shows
-    residuals about zero -- a figure is still produced, faithful to R's colors.
+def _render_m_reg(fitted: dict[str, NDArray[np.float64] | NDArray[np.str_]]) -> None:
+    """Render R's NNS.M.reg residual plot (Multivariate_Regression.R:367-377).
+
+    The multivariate plot output is the residual plot: actual ``y`` over the
+    observation index as ``steelblue`` open circles, fitted ``y.hat`` as a
+    ``red`` line, and a pink (alpha 0.375) confidence band when present.
     """
+    from nns.plotting import palette
     from nns.plotting._mpl import resolve_ax
 
     y = np.asarray(fitted["y"], dtype=np.float64)
     y_hat = np.asarray(fitted["y.hat"], dtype=np.float64)
-    if plot and y.size and y.size == y_hat.size:
-        ax = resolve_ax(None)
-        ax.scatter(y, y_hat, color="steelblue")
-        lo, hi = float(min(y.min(), y_hat.min())), float(max(y.max(), y_hat.max()))
-        ax.plot([lo, hi], [lo, hi], color="red")
-        ax.set_xlabel("y")
-        ax.set_ylabel("y.hat")
-        ax.set_title("NNS.M.reg Fitted vs Actual")
-    if residual_plot:
-        residuals = np.asarray(fitted.get("residuals", []), dtype=np.float64)
-        if residuals.size:
-            ax = resolve_ax(None)
-            ax.scatter(np.arange(1, residuals.size + 1), residuals, color="steelblue")
-            ax.axhline(0.0, color="red")
-            ax.set_title("NNS.M.reg Residual Plot")
+    if y.size == 0 or y.size != y_hat.size:
+        return
+    ax = resolve_ax(None)
+    index = np.arange(1, y.size + 1)
+    ax.scatter(index, y, facecolors="none", edgecolors="steelblue", marker="o")
+    ax.plot(index, y_hat, color="red", linewidth=2)
+    if "conf.int.pos" in fitted and "conf.int.neg" in fitted:
+        pos = np.asarray(fitted["conf.int.pos"], dtype=np.float64)
+        neg = np.asarray(fitted["conf.int.neg"], dtype=np.float64)
+        mask = np.isfinite(pos) & np.isfinite(neg)
+        if mask.any():
+            ax.fill_between(
+                index[mask], neg[mask], pos[mask],
+                color=palette.PINK, alpha=palette.CI_ALPHA_REG, linewidth=0.0,
+            )
+    ax.set_xlabel("Index")
+    ax.set_ylabel("y (blue)   y.hat (red)")
+    ax.set_title("NNS.M.reg Residual Plot")
 
 
 def _validate_inputs(

tests/plotting/test_compute_plot_flag.py

@@ -12,6 +12,8 @@
 
 matplotlib.use("Agg")
 
+from typing import Any
+
 import matplotlib.pyplot as plt
 import numpy as np
 import pytest
@@ -88,12 +90,25 @@ def test_nns_seas_plot_true_creates_figure() -> None:
 
 
 def test_nns_m_reg_plot_true_creates_figure() -> None:
+    import matplotlib.colors as mcolors
+
     rng = np.random.default_rng(6)
     x = np.sort(rng.normal(size=40))
     y = 2.0 * x + rng.normal(scale=0.3, size=40)
     features = np.column_stack([x, x**2])
     nns.nns_m_reg(features, y, plot=True)
     assert _fig_count() > 0
+    ax: Any = plt.gca()
+    # R's M.reg residual plot: actual y is steelblue, fitted y.hat is a red line.
+    edge_hexes = {
+        mcolors.to_hex(row)
+        for coll in ax.collections
+        for row in coll.get_edgecolor()
+        if len(row)
+    }
+    line_hexes = {mcolors.to_hex(line.get_color()) for line in ax.get_lines()}
+    assert "#4682b4" in edge_hexes
+    assert "#ff0000" in line_hexes
 
 
 def test_plot_false_creates_no_figure() -> None: