openstat-cli 1.0.0__py3-none-any.whl

openstat/commands/sem_cmds.py

@@ -0,0 +1,177 @@
+"""Structural Equation Modeling (SEM) commands."""
+
+from __future__ import annotations
+
+from openstat.commands.base import command
+from openstat.session import Session
+
+
+def _fmt_sem(title: str, lines_body: list[str]) -> str:
+    sep = "=" * 60
+    return "\n".join([f"\n{title}", sep] + lines_body + [sep])
+
+
+@command("sem", usage="sem model_syntax [--method=ml|gls|wls]")
+def cmd_sem(session: Session, args: str) -> str:
+    """Structural Equation Model estimation via semopy.
+
+    Model syntax (lavaan/semopy style):
+      Latent variable:   F1 =~ x1 + x2 + x3
+      Regression:        y ~ x1 + x2
+      Covariance:        x1 ~~ x2
+
+    Examples:
+      sem 'F1 =~ age + income + score'
+      sem 'eta =~ x1 + x2 + x3 \\n y ~ eta + z'
+      sem 'F1 =~ x1 + x2 \\n y ~ F1' --method=ml
+
+    Requires: pip install semopy
+    """
+    try:
+        import semopy
+    except ImportError:
+        return "SEM requires semopy. Install with: pip install semopy"
+
+    df = session.require_data()
+
+    import re
+    args = args.strip()
+
+    # Parse --method option
+    method = "ml"
+    m = re.search(r"--method[= ](\w+)", args)
+    if m:
+        method = m.group(1).lower()
+        args = re.sub(r"--method[= ]\w+", "", args).strip()
+
+    # Strip surrounding quotes
+    model_str = args.strip("\"'")
+    # Support \n in quoted strings as actual newlines
+    model_str = model_str.replace("\\n", "\n")
+
+    if not model_str:
+        return (
+            "Usage: sem '<model_syntax>' [--method=ml|gls|wls]\n"
+            "Example: sem 'F1 =~ x1 + x2 + x3'\n"
+            "Example: sem 'y ~ x1 + x2'"
+        )
+
+    # Map user-friendly method names to semopy obj parameter
+    _method_map = {"ml": "MLW", "gls": "GLS", "wls": "WLS", "uls": "ULS", "fiml": "FIML"}
+    obj = _method_map.get(method.lower(), "MLW")
+
+    try:
+        pandas_df = df.to_pandas()
+        model = semopy.Model(model_str)
+        res = model.fit(pandas_df, obj=obj)
+
+        # Model fit indices
+        stats = semopy.calc_stats(model)
+
+        lines = [
+            f"Model: {model_str.replace(chr(10), ' | ')}",
+            f"Method: {method.upper()}   N = {len(pandas_df)}",
+            "",
+        ]
+
+        # Parameter estimates
+        try:
+            params = model.inspect()
+            lines.append("Parameter Estimates:")
+            lines.append(f"  {'lval':<12} {'op':<5} {'rval':<12} {'Estimate':>10} {'Std.Err':>9} {'z':>7} {'p-value':>9}")
+            lines.append("  " + "-" * 65)
+            for _, row in params.iterrows():
+                p_str = f"{row.get('p-value', float('nan')):.4f}" if row.get('p-value') is not None else "   ."
+                z_str = f"{row.get('z-value', float('nan')):.3f}" if row.get('z-value') is not None else "   ."
+                se_str = f"{row.get('Std. Err', float('nan')):.4f}" if row.get('Std. Err') is not None else "   ."
+                lines.append(
+                    f"  {str(row.get('lval','')):<12} {str(row.get('op','')):<5} "
+                    f"{str(row.get('rval','')):<12} "
+                    f"{row.get('Estimate', float('nan')):>10.4f} "
+                    f"{se_str:>9} {z_str:>7} {p_str:>9}"
+                )
+        except Exception:
+            lines.append("(Parameter table unavailable)")
+
+        # Fit indices
+        lines.append("")
+        lines.append("Model Fit Indices:")
+        fit_map = {
+            "CFI": "CFI", "GFI": "GFI", "AGFI": "AGFI",
+            "NFI": "NFI", "RMSEA": "RMSEA", "SRMR": "SRMR",
+            "chi2": "Chi-sq", "DoF": "df", "pvalue": "p(chi-sq)",
+            "AIC": "AIC", "BIC": "BIC",
+        }
+        for key, label in fit_map.items():
+            try:
+                val = stats[key].iloc[0] if hasattr(stats[key], 'iloc') else stats.get(key)
+                if val is not None:
+                    lines.append(f"  {label:<12} {float(val):.4f}")
+            except Exception:
+                pass
+
+        # Interpretation hints
+        lines.append("")
+        lines.append("Fit guidelines: CFI/GFI > 0.95 (good), RMSEA < 0.05 (good), SRMR < 0.08 (good)")
+
+        return _fmt_sem("Structural Equation Model (SEM)", lines)
+
+    except Exception as exc:
+        return f"SEM error: {exc}"
+
+
+@command("cfa", usage="cfa factor =~ x1 + x2 + x3 [+ factor2 =~ y1 + y2]")
+def cmd_cfa(session: Session, args: str) -> str:
+    """Confirmatory Factor Analysis — shortcut for SEM with =~ syntax only.
+
+    Examples:
+      cfa F1 =~ age + income + score
+      cfa 'F1 =~ x1 + x2 \\n F2 =~ y1 + y2'
+    """
+    try:
+        import semopy
+    except ImportError:
+        return "CFA requires semopy. Install with: pip install semopy"
+
+    df = session.require_data()
+    model_str = args.strip().strip("\"'").replace("\\n", "\n")
+    if not model_str:
+        return "Usage: cfa 'F1 =~ x1 + x2 + x3'"
+
+    try:
+        pandas_df = df.to_pandas()
+        model = semopy.Model(model_str)
+        model.fit(pandas_df)
+
+        params = model.inspect()  # type: ignore[attr-defined]
+        stats = semopy.calc_stats(model)
+
+        lines = [f"CFA Model: {model_str.replace(chr(10), ' | ')}", f"N = {len(pandas_df)}", ""]
+
+        # Loadings only (=~ relationships)
+        loadings = params[params["op"] == "=~"] if "op" in params.columns else params
+        if not loadings.empty:
+            lines.append("Factor Loadings:")
+            lines.append(f"  {'Factor':<12} {'Indicator':<12} {'Loading':>9} {'Std.Err':>9} {'p-value':>9}")
+            lines.append("  " + "-" * 55)
+            for _, row in loadings.iterrows():
+                p_str = f"{row.get('p-value', float('nan')):.4f}"
+                se_str = f"{row.get('Std. Err', float('nan')):.4f}"
+                lines.append(
+                    f"  {str(row.get('lval','')):<12} {str(row.get('rval','')):<12} "
+                    f"{row.get('Estimate', float('nan')):>9.4f} {se_str:>9} {p_str:>9}"
+                )
+
+        lines.append("")
+        lines.append("Fit Indices:")
+        for key, label in [("CFI", "CFI"), ("RMSEA", "RMSEA"), ("SRMR", "SRMR"), ("AIC", "AIC")]:
+            try:
+                val = stats[key].iloc[0]
+                lines.append(f"  {label:<10} {float(val):.4f}")
+            except Exception:
+                pass
+
+        return _fmt_sem("Confirmatory Factor Analysis (CFA)", lines)
+
+    except Exception as exc:
+        return f"CFA error: {exc}"