dasycaus 1.0.0__tar.gz → 1.0.1__tar.gz

{dasycaus-1.0.0/dasycaus.egg-info → dasycaus-1.0.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dasycaus
-Version: 1.0.0
+Version: 1.0.1
 Summary: Dynamic Asymmetric Causality Tests for Time Series Analysis
 Author-email: "Dr. Merwan Roudane" <merwanroudane920@gmail.com>
 Maintainer-email: "Dr. Merwan Roudane" <merwanroudane920@gmail.com>

{dasycaus-1.0.0 → dasycaus-1.0.1}/dasycaus/__init__.py

@@ -21,7 +21,7 @@ Based on:
   Journal of Economic Studies, 39(2), 144-160.
 """
 
-__version__ = "1.0.0"
+__version__ = "1.0.1"
 __author__ = "Dr. Merwan Roudane"
 __email__ = "merwanroudane920@gmail.com"
 
@@ -53,6 +53,14 @@ from .utils import (
     plot_dynamic_causality
 )
 
+from .diagnostics import (
+    doornik_hansen_test,
+    multivariate_arch_test,
+    multivariate_autocorrelation_test,
+    run_all_diagnostics,
+    print_diagnostic_results
+)
+
 __all__ = [
     'dynamic_asymmetric_causality_test',
     'dynamic_symmetric_causality_test',
@@ -66,5 +74,10 @@ __all__ = [
     'information_criteria',
     'generate_subsamples',
     'compute_test_ratio',
-    'plot_dynamic_causality'
+    'plot_dynamic_causality',
+    'doornik_hansen_test',
+    'multivariate_arch_test',
+    'multivariate_autocorrelation_test',
+    'run_all_diagnostics',
+    'print_diagnostic_results'
 ]

{dasycaus-1.0.0 → dasycaus-1.0.1}/dasycaus/bootstrap.py

@@ -13,7 +13,7 @@ def bootstrap_critical_values(
     data: np.ndarray,
     lag_order: int,
     integration_order: int,
-    num_simulations: int = 1000,
+    num_simulations: int = 10000,
     significance_level: float = 0.05,
     random_seed: Optional[int] = None,
     leverage_adjustment: bool = True
@@ -29,7 +29,7 @@ def bootstrap_critical_values(
         Lag order for VAR model.
     integration_order : int
         Integration order (for adding extra lags).
-    num_simulations : int, default=1000
+    num_simulations : int, default=10000
         Number of bootstrap simulations.
     significance_level : float, default=0.05
         Significance level for critical value.
@@ -94,7 +94,7 @@ def leveraged_bootstrap(
     data: np.ndarray,
     lag_order: int,
     integration_order: int,
-    num_simulations: int = 1000,
+    num_simulations: int = 10000,
     random_seed: Optional[int] = None
 ) -> np.ndarray:
     """
@@ -108,7 +108,7 @@ def leveraged_bootstrap(
         Lag order for VAR model.
     integration_order : int
         Integration order.
-    num_simulations : int, default=1000
+    num_simulations : int, default=10000
         Number of bootstrap simulations.
     random_seed : int, optional
         Random seed for reproducibility.

{dasycaus-1.0.0 → dasycaus-1.0.1}/dasycaus/core.py

@@ -19,7 +19,7 @@ def symmetric_causality_test(
     maxlags: int,
     integration_order: int = 1,
     info_criterion: str = 'hjc',
-    bootstrap_sims: int = 1000,
+    bootstrap_sims: int = 10000,
     significance_levels: List[float] = [0.05, 0.10],
     random_seed: Optional[int] = 30540
 ) -> Dict:
@@ -37,7 +37,7 @@ def symmetric_causality_test(
         Integration order of the variables (0=stationary, 1=I(1), 2=I(2)).
     info_criterion : str, default='hjc'
         Information criterion for lag selection: 'aic', 'aicc', 'sbc', 'hqc', 'hjc'.
-    bootstrap_sims : int, default=1000
+    bootstrap_sims : int, default=10000
         Number of bootstrap simulations for critical values.
     significance_levels : list, default=[0.05, 0.10]
         List of significance levels for testing.
@@ -113,10 +113,10 @@ def asymmetric_causality_test(
     component: str = 'positive',
     integration_order: int = 1,
     info_criterion: str = 'hjc',
-    bootstrap_sims: int = 1000,
+    bootstrap_sims: int = 10000,
     significance_levels: List[float] = [0.05, 0.10],
     random_seed: Optional[int] = 30540,
-    include_trend: bool = True
+    include_trend: bool = False
 ) -> Dict:
     """
     Conduct static asymmetric Granger causality test.
@@ -133,14 +133,16 @@ def asymmetric_causality_test(
         Integration order of the original variables.
     info_criterion : str, default='hjc'
         Information criterion for lag selection.
-    bootstrap_sims : int, default=1000
+    bootstrap_sims : int, default=10000
         Number of bootstrap simulations.
     significance_levels : list, default=[0.05, 0.10]
         List of significance levels.
     random_seed : int, optional
         Random seed for reproducibility.
-    include_trend : bool, default=True
+    include_trend : bool, default=False
         Whether to include deterministic trend in transformation.
+        Default False matches GAUSS implementation (simple cumulative sums).
+        Set to True for paper equations (5-8) with deterministic trends.
         
     Returns
     -------
@@ -180,7 +182,7 @@ def dynamic_symmetric_causality_test(
     maxlags: int,
     integration_order: int = 1,
     info_criterion: str = 'hjc',
-    bootstrap_sims: int = 1000,
+    bootstrap_sims: int = 10000,
     significance_levels: List[float] = [0.05, 0.10],
     subsample_method: str = 'recursive',
     random_seed: Optional[int] = 30540
@@ -198,7 +200,7 @@ def dynamic_symmetric_causality_test(
         Integration order of the variables.
     info_criterion : str, default='hjc'
         Information criterion for lag selection.
-    bootstrap_sims : int, default=1000
+    bootstrap_sims : int, default=10000
         Number of bootstrap simulations.
     significance_levels : list, default=[0.05, 0.10]
         List of significance levels.
@@ -280,11 +282,11 @@ def dynamic_asymmetric_causality_test(
     component: str = 'positive',
     integration_order: int = 1,
     info_criterion: str = 'hjc',
-    bootstrap_sims: int = 1000,
+    bootstrap_sims: int = 10000,
     significance_levels: List[float] = [0.05, 0.10],
     subsample_method: str = 'recursive',
     random_seed: Optional[int] = 30540,
-    include_trend: bool = True
+    include_trend: bool = False
 ) -> Dict:
     """
     Conduct dynamic asymmetric Granger causality tests using subsamples.
@@ -301,7 +303,7 @@ def dynamic_asymmetric_causality_test(
         Integration order of the original variables.
     info_criterion : str, default='hjc'
         Information criterion for lag selection.
-    bootstrap_sims : int, default=1000
+    bootstrap_sims : int, default=10000
         Number of bootstrap simulations.
     significance_levels : list, default=[0.05, 0.10]
         List of significance levels.
@@ -309,8 +311,9 @@ def dynamic_asymmetric_causality_test(
         Method for creating subsamples: 'recursive' or 'rolling'.
     random_seed : int, optional
         Random seed for reproducibility.
-    include_trend : bool, default=True
+    include_trend : bool, default=False
         Whether to include deterministic trend in transformation.
+        Default False matches GAUSS implementation.
         
     Returns
     -------
@@ -378,7 +381,9 @@ def _compute_wald_statistic(
     # Estimate unrestricted VAR
     A_hat = np.linalg.lstsq(X, Y, rcond=None)[0]
     residuals = Y - X @ A_hat
-    Sigma_u = (residuals.T @ residuals) / (T - total_lags - 1)
+    T_eff = Y.shape[0]  # Effective sample size after lagging
+    # Use T_eff as denominator (matches GAUSS line 866: VARCOV = RES'RES/T)
+    Sigma_u = (residuals.T @ residuals) / T_eff
     
     # Create restriction matrix C
     # We test if variable 2 does not cause variable 1

dasycaus-1.0.1/dasycaus/diagnostics.py

@@ -0,0 +1,395 @@
+"""
+Diagnostic tests module for VAR residuals.
+
+This module implements diagnostic tests for VAR model residuals as used in
+Hatemi-J's causality testing framework:
+- Multivariate Normality Test (Doornik-Hansen, 2008)
+- Multivariate ARCH Test (Hacker-Hatemi-J, 2005)
+- Multivariate Autocorrelation Test (Hatemi-J, 2004)
+"""
+
+import numpy as np
+from typing import Dict, Optional, Tuple
+from scipy.stats import chi2
+
+
+def doornik_hansen_test(residuals: np.ndarray) -> Dict:
+    """
+    Multivariate normality test (Doornik and Hansen, 2008).
+    
+    Tests the null hypothesis that residuals follow a multivariate normal distribution.
+    
+    Parameters
+    ----------
+    residuals : np.ndarray
+        Residuals matrix with shape (T, n) where T is number of observations
+        and n is number of variables.
+        
+    Returns
+    -------
+    dict
+        Dictionary containing:
+        - 'test_statistic': Overall test statistic
+        - 'p_value': P-value for the test
+        - 'df': Degrees of freedom
+        - 'reject_null_5pct': Boolean indicating rejection at 5% level
+        - 'individual_stats': Dict with per-variable statistics
+        
+    References
+    ----------
+    Doornik, J.A. and Hansen, H. (2008). An omnibus test for univariate and
+    multivariate normality. Oxford Bulletin of Economics and Statistics, 70, 927-939.
+    """
+    T, n = residuals.shape
+    
+    if T < 4:
+        raise ValueError("Need at least 4 observations for normality test")
+    
+    # Standardize residuals
+    means = np.mean(residuals, axis=0)
+    stds = np.std(residuals, axis=0, ddof=1)
+    z = (residuals - means) / stds
+    
+    # Compute skewness and kurtosis for each variable
+    skewness = np.zeros(n)
+    kurtosis = np.zeros(n)
+    
+    for i in range(n):
+        m3 = np.mean(z[:, i]**3)
+        m4 = np.mean(z[:, i]**4)
+        skewness[i] = m3
+        kurtosis[i] = m4 - 3  # Excess kurtosis
+    
+    # Transformation for skewness
+    beta = (3 * (T**2 + 27*T - 70) * (T+1) * (T+3)) / ((T-2) * (T+5) * (T+7) * (T+9))
+    w2 = -1 + np.sqrt(2 * (beta - 1))
+    delta = 1 / np.sqrt(np.log(w2))
+    alpha = np.sqrt(2 / (w2 - 1))
+    
+    z_skew = np.zeros(n)
+    for i in range(n):
+        y = skewness[i] * alpha
+        z_skew[i] = delta * np.log(y + np.sqrt(y**2 + 1))
+    
+    # Transformation for kurtosis
+    delta_k = (T - 3) * (T + 1) * (T**2 + 15*T - 4)
+    a = ((T-2) * (T+5) * (T+7) * (T**2 + 27*T - 70)) / (6 * delta_k)
+    c = ((T-7) * (T+5) * (T+7) * (T**2 + 2*T - 5)) / (6 * delta_k)
+    k = ((T+5) * (T+7) * (T**3 + 37*T**2 + 11*T - 313)) / (12 * delta_k)
+    
+    z_kurt = np.zeros(n)
+    for i in range(n):
+        x = kurtosis[i] / np.sqrt(8 / T)
+        z_kurt[i] = (np.power(1 - 2*c, 0.5) / np.sqrt(2*a)) * \
+                    (np.power(x / (2*a) + 1 + k, 1/3) - 1)
+    
+    # Combine statistics
+    individual_stats = {}
+    omnibus_stat = 0
+    
+    for i in range(n):
+        var_stat = z_skew[i]**2 + z_kurt[i]**2
+        omnibus_stat += var_stat
+        individual_stats[f'variable_{i+1}'] = {
+            'statistic': var_stat,
+            'skewness': skewness[i],
+            'kurtosis': kurtosis[i]
+        }
+    
+    # Degrees of freedom: 2 * n (2 per variable for skewness and kurtosis)
+    df = 2 * n
+    p_value = 1 - chi2.cdf(omnibus_stat, df)
+    
+    return {
+        'test_statistic': float(omnibus_stat),
+        'p_value': float(p_value),
+        'df': df,
+        'reject_null_5pct': p_value < 0.05,
+        'individual_stats': individual_stats,
+        'test_name': 'Doornik-Hansen Multivariate Normality Test'
+    }
+
+
+def multivariate_arch_test(
+    residuals: np.ndarray,
+    lags: int = 1
+) -> Dict:
+    """
+    Multivariate ARCH test (Hacker and Hatemi-J, 2005).
+    
+    Tests the null hypothesis of no ARCH effects in VAR residuals.
+    
+    Parameters
+    ----------
+    residuals : np.ndarray
+        Residuals matrix with shape (T, n).
+    lags : int, default=1
+        Number of lags to test for ARCH effects.
+        
+    Returns
+    -------
+    dict
+        Dictionary containing:
+        - 'test_statistic': Test statistic
+        - 'p_value': P-value for the test
+        - 'df': Degrees of freedom
+        - 'reject_null_5pct': Boolean indicating rejection at 5% level
+        
+    References
+    ----------
+    Hacker, S. and Hatemi-J, A. (2005). A multivariate test for ARCH effects.
+    Applied Economics Letters, 12(7), 411-417.
+    """
+    T, n = residuals.shape
+    
+    if T <= lags:
+        raise ValueError(f"Need more than {lags} observations for ARCH test")
+    
+    # Compute squared residuals
+    e_squared = residuals ** 2
+    
+    # Stack into vector
+    e_vec = e_squared.flatten('F')  # Column-major order
+    
+    # Create lagged matrix for regression
+    Y = e_vec[n*lags:]  # Dependent variable
+    
+    # Create X matrix with lags
+    X_list = []
+    for lag in range(1, lags + 1):
+        start_idx = n*(lags - lag)
+        end_idx = start_idx + (T - lags)*n
+        X_list.append(e_vec[start_idx:end_idx].reshape(-1, 1))
+    
+    if not X_list:
+        raise ValueError("Could not create lagged matrix")
+    
+    X = np.hstack(X_list)
+    X = np.column_stack([np.ones(len(Y)), X])  # Add intercept
+    
+    # OLS regression
+    try:
+        beta_hat = np.linalg.lstsq(X, Y, rcond=None)[0]
+        residuals_ols = Y - X @ beta_hat
+        SSR = np.sum(residuals_ols ** 2)
+        SST = np.sum((Y - np.mean(Y)) ** 2)
+        R_squared = 1 - SSR / SST
+    except np.linalg.LinAlgError:
+        # If singular matrix, return conservative result
+        return {
+            'test_statistic': 0.0,
+            'p_value': 1.0,
+            'df': n * lags,
+            'reject_null_5pct': False,
+            'test_name': 'Multivariate ARCH Test',
+            'warning': 'Singular matrix encountered'
+        }
+    
+    # Test statistic: T * R^2
+    test_stat = (T - lags) * R_squared
+    df = n * lags
+    p_value = 1 - chi2.cdf(test_stat, df)
+    
+    return {
+        'test_statistic': float(test_stat),
+        'p_value': float(p_value),
+        'df': df,
+        'reject_null_5pct': p_value < 0.05,
+        'R_squared': float(R_squared),
+        'test_name': 'Multivariate ARCH Test'
+    }
+
+
+def multivariate_autocorrelation_test(
+    residuals: np.ndarray,
+    lags: int
+) -> Dict:
+    """
+    Multivariate autocorrelation test (Hatemi-J, 2004).
+    
+    Tests the null hypothesis of no autocorrelation in VAR residuals.
+    
+    Parameters
+    ----------
+    residuals : np.ndarray
+        Residuals matrix with shape (T, n).
+    lags : int
+        Number of lags to test for autocorrelation.
+        
+    Returns
+    -------
+    dict
+        Dictionary containing:
+        - 'test_statistic': Test statistic
+        - 'p_value': P-value for the test
+        - 'df': Degrees of freedom
+        - 'reject_null_5pct': Boolean indicating rejection at 5% level
+        
+    References
+    ----------
+    Hatemi-J, A. (2004). Multivariate tests for autocorrelation in the stable
+    and unstable VAR models. Economic Modelling, 21, 661-683.
+    """
+    T, n = residuals.shape
+    
+    if T <= lags:
+        raise ValueError(f"Need more than {lags} observations for autocorrelation test")
+    
+    # Compute residual covariance matrix
+    Sigma_0 = (residuals.T @ residuals) / T
+    
+    # Compute autocov matrices and test statistic
+    test_stat = 0
+    
+    for h in range(1, lags + 1):
+        # Autocovariance at lag h
+        Gamma_h = np.zeros((n, n))
+        for t in range(h, T):
+            Gamma_h += np.outer(residuals[t], residuals[t - h])
+        Gamma_h /= T
+        
+        # Correlation matrix at lag h
+        try:
+            Sigma_0_inv = np.linalg.inv(Sigma_0)
+            D_h = Gamma_h @ Sigma_0_inv
+            
+            # Add to test statistic
+            test_stat += T * np.trace(D_h.T @ D_h)
+        except np.linalg.LinAlgError:
+            # If singular, skip this lag
+            continue
+    
+    # Degrees of freedom
+    df = n * n * lags
+    p_value = 1 - chi2.cdf(test_stat, df)
+    
+    return {
+        'test_statistic': float(test_stat),
+        'p_value': float(p_value),
+        'df': df,
+        'reject_null_5pct': p_value < 0.05,
+        'test_name': 'Multivariate Autocorrelation Test'
+    }
+
+
+def run_all_diagnostics(
+    residuals: np.ndarray,
+    arch_lags: int = 1,
+    autocorr_lags: Optional[int] = None
+) -> Dict:
+    """
+    Run all diagnostic tests on VAR residuals.
+    
+    Parameters
+    ----------
+    residuals : np.ndarray
+        Residuals matrix with shape (T, n).
+    arch_lags : int, default=1
+        Number of lags for ARCH test.
+    autocorr_lags : int, optional
+        Number of lags for autocorrelation test. If None, uses arch_lags.
+        
+    Returns
+    -------
+    dict
+        Dictionary containing results from all tests:
+        - 'normality': Results from Doornik-Hansen test
+        - 'arch': Results from multivariate ARCH test
+        - 'autocorrelation': Results from autocorrelation test
+        - 'summary': Overall summary
+    """
+    if autocorr_lags is None:
+        autocorr_lags = arch_lags
+    
+    results = {}
+    
+    # Run normality test
+    try:
+        results['normality'] = doornik_hansen_test(residuals)
+    except Exception as e:
+        results['normality'] = {
+            'error': str(e),
+            'test_name': 'Doornik-Hansen Multivariate Normality Test'
+        }
+    
+    # Run ARCH test
+    try:
+        results['arch'] = multivariate_arch_test(residuals, arch_lags)
+    except Exception as e:
+        results['arch'] = {
+            'error': str(e),
+            'test_name': 'Multivariate ARCH Test'
+        }
+    
+    # Run autocorrelation test
+    try:
+        results['autocorrelation'] = multivariate_autocorrelation_test(
+            residuals, autocorr_lags
+        )
+    except Exception as e:
+        results['autocorrelation'] = {
+            'error': str(e),
+            'test_name': 'Multivariate Autocorrelation Test'
+        }
+    
+    # Create summary
+    summary = []
+    for test_name, test_result in results.items():
+        if 'error' in test_result:
+            summary.append(f"{test_name}: ERROR - {test_result['error']}")
+        else:
+            reject = test_result.get('reject_null_5pct', False)
+            p_val = test_result.get('p_value', np.nan)
+            summary.append(
+                f"{test_name}: p-value={p_val:.4f}, "
+                f"reject_5%={reject}"
+            )
+    
+    results['summary'] = '\n'.join(summary)
+    
+    return results
+
+
+def print_diagnostic_results(results: Dict) -> None:
+    """
+    Print diagnostic test results in a formatted table.
+    
+    Parameters
+    ----------
+    results : dict
+        Results from run_all_diagnostics function.
+    """
+    print("\n" + "="*70)
+    print("DIAGNOSTIC TESTS FOR VAR RESIDUALS")
+    print("="*70)
+    
+    for test_key in ['normality', 'arch', 'autocorrelation']:
+        if test_key not in results:
+            continue
+            
+        test_result = results[test_key]
+        test_name = test_result.get('test_name', test_key.upper())
+        
+        print(f"\n{test_name}:")
+        print("-" * 70)
+        
+        if 'error' in test_result:
+            print(f"  ERROR: {test_result['error']}")
+            continue
+        
+        print(f"  Test Statistic: {test_result['test_statistic']:.4f}")
+        print(f"  Degrees of Freedom: {test_result['df']}")
+        print(f"  P-value: {test_result['p_value']:.4f}")
+        print(f"  Reject H0 at 5%: {test_result['reject_null_5pct']}")
+        
+        # Additional info for specific tests
+        if test_key == 'arch' and 'R_squared' in test_result:
+            print(f"  R-squared: {test_result['R_squared']:.4f}")
+    
+    print("\n" + "="*70)
+    print("INTERPRETATION:")
+    print("-" * 70)
+    print("Normality: H0 = Residuals are multivariate normal")
+    print("ARCH: H0 = No ARCH effects (constant variance)")
+    print("Autocorrelation: H0 = No autocorrelation")
+    print("="*70 + "\n")

{dasycaus-1.0.0 → dasycaus-1.0.1/dasycaus.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dasycaus
-Version: 1.0.0
+Version: 1.0.1
 Summary: Dynamic Asymmetric Causality Tests for Time Series Analysis
 Author-email: "Dr. Merwan Roudane" <merwanroudane920@gmail.com>
 Maintainer-email: "Dr. Merwan Roudane" <merwanroudane920@gmail.com>

{dasycaus-1.0.0 → dasycaus-1.0.1}/dasycaus.egg-info/SOURCES.txt

@@ -7,6 +7,7 @@ dasycaus/__init__.py
 dasycaus/bootstrap.py
 dasycaus/core.py
 dasycaus/data_transform.py
+dasycaus/diagnostics.py
 dasycaus/lag_selection.py
 dasycaus/utils.py
 dasycaus.egg-info/PKG-INFO

{dasycaus-1.0.0 → dasycaus-1.0.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "dasycaus"
-version = "1.0.0"
+version = "1.0.1"
 description = "Dynamic Asymmetric Causality Tests for Time Series Analysis"
 readme = "README.md"
 authors = [