econmethods 2.8__tar.gz → 3__tar.gz

{econmethods-2.8 → econmethods-3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: econmethods
-Version: 2.8
+Version: 3
 Summary: A python package implementing various econometrical tests and estimators
 Home-page: https://github.com/NaturionBG/econmethods
 Author: NaturionBG

{econmethods-2.8 → econmethods-3}/econmethods/Lib.py

@@ -278,7 +278,6 @@ class HausmanOneWay:
   def __del__(self) -> None:
     pass
 
-
 class FECM:
   '''
   The implementation of a first-order ECM (Error Correction Model) estimation for panel data.
@@ -312,6 +311,7 @@ class FECM:
   - *intercept*: Specify whether the ECM model should have an intercept. Defaults to True.
   - *stat_vars*: a DataFrame of the same format as "df" - includes variables that will not be differenced and included into the ECM in their raw form. Ensure these variables are I(0). Defaults to None.
   - *lr_const*: Specify whether the long-run model should have a constant.
+  - *autorem*: Enables/Disables automatic removal of insignificant regression coefficients in the ECM. Defaults to True.
   ----
   RETURNS:
   --
@@ -321,12 +321,14 @@ class FECM:
     If a CCE- method is chosen:
     - The AR(d) estimation results to forecast the cross-sectional mean | key = "ar"
   '''
-  def __init__(self, df: pd.DataFrame, effects: str = 'rand', trend: int = 0, n_lags: int = 1, method: str = 'MG', coint: str | list[str] = 'x1', intercept: bool = True, stat_vars: pd.DataFrame|None = None, lr_const: bool = False) -> None:
+  def __init__(self, df: pd.DataFrame, effects: str = 'rand', trend: int = 0, n_lags: int = 1, method: str = 'MG', coint: str | list[str] = 'x1', intercept: bool = True, stat_vars: pd.DataFrame|None = None, lr_const: bool = False, autorem: bool = True) -> None:
     self.__df = df
     self.__eff = effects.lower()
     self.__t = trend
+    self.__autorem = autorem
     self.__C = intercept
     self.__lag = n_lags
+    self.x_difs = []
     self.__method = method.lower()
     self.__exog = len(df.columns[3:])
     self.__l =[]
@@ -340,6 +342,7 @@ class FECM:
       self.__stat_vars.columns = ['SpUnit', 'time'] + self.__stat
     for i in range(1, self.__exog+1):
       self.__l.append(f'x{i}')
+      self.x_difs.append(f'x{i}_diff')
       self.__mean_names.append(f'x{i}_avg')
     self.__df.columns = ['SpUnit', 'time', 'target'] + self.__l
     if isinstance(coint, list):
@@ -510,18 +513,28 @@ class FECM:
         est.append(sm.OLS(pool['target_diff'], sm.add_constant(pool.iloc[:, 3:])).fit())
       else:
         est.append(sm.OLS(pool['target_diff'], pool.iloc[:, 3:]).fit())
-      while True:
-        flag = True
-        if max(zip(est[0].params.index, est[0].pvalues), key=lambda x: x[1])[1] > 0.06:
-          pool = pool.drop(columns=[max(zip(est[0].params.index, est[0].pvalues), key=lambda x: x[1])[0]])
-          flag=False
-        if flag:
-          break
-        else:
-          if self.__C:
-            est[0] = sm.OLS(pool['target_diff'], sm.add_constant(pool.iloc[:, 3:])).fit()
+      if self.__autorem:
+        while True:
+          ts = []
+          flag = True
+          if max(zip(est[0].params.index, est[0].pvalues), key=lambda x: x[1])[1] > 0.06:
+            pool = pool.drop(columns=[max(zip(est[0].params.index, est[0].pvalues), key=lambda x: x[1])[0]])
+            flag=False
+          for i in self.x_difs:
+            try:
+              tmp = est[0].pvalues[i] <= 0.05
+              if tmp:
+                ts.append(1)
+            except Exception:
+              ts.append(1)
+              
+          if flag or len(ts) == len(self.x_difs):
+            break
           else:
-            est[0] = sm.OLS(pool['target_diff'], pool.iloc[:, 3:]).fit()
+            if self.__C:
+              est[0] = sm.OLS(pool['target_diff'], sm.add_constant(pool.iloc[:, 3:])).fit()
+            else:
+              est[0] = sm.OLS(pool['target_diff'], pool.iloc[:, 3:]).fit()
       return est
 
   def mg_algorithm(self) -> pd.DataFrame:
@@ -554,25 +567,30 @@ class FECM:
     elif self.__method == 'ccemg' or self.__method == 'mg':
       dct['lr_res'] = self.__lr
       res = self.mg_algorithm()
-      while True:
-        flag = True
-        if max(zip(res['coefs'].index, res['coefs']['T-pvalues']), key=lambda x: x[1])[1] > 0.06:
-          self.__sr = self.build_sr([max(zip(res['coefs'].index, res['coefs']['T-pvalues']), key=lambda x: x[1])[0]])
-          flag=False
-        if flag:
-          break
-        else:
-          res =  self.mg_algorithm()
+      if self.__autorem:
+        while True:
+          ts = []
+          flag = True
+          if max(zip(res['coefs'].index, res['coefs']['T-pvalues']), key=lambda x: x[1])[1] > 0.06:
+            self.__sr = self.build_sr([max(zip(res['coefs'].index, res['coefs']['T-pvalues']), key=lambda x: x[1])[0]])
+            flag=False
+          for i in self.x_difs:
+            try:
+              tmp = res['coefs']['T-pvalues'][i] <= 0.05
+              if tmp:
+                ts.append(1)
+            except Exception:
+              ts.append(1)
+          if flag or len(ts) == len(self.x_difs):
+            break
+          else:
+            res =  self.mg_algorithm()
 
       dct['sr_res'] = res
     return dct
   
   def __del__(self) -> None:
     pass
-
-
-
-
 class CDTwoWay:
   '''
   Implementation of the CD test to validate/reject cross-sectional dependence.

{econmethods-2.8 → econmethods-3}/econmethods.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: econmethods
-Version: 2.8
+Version: 3
 Summary: A python package implementing various econometrical tests and estimators
 Home-page: https://github.com/NaturionBG/econmethods
 Author: NaturionBG

{econmethods-2.8 → econmethods-3}/setup.py

@@ -5,7 +5,7 @@ with open('README.md', 'r') as f:
   
 setup(
   name = 'econmethods',
-  version = 'v2.8',
+  version = 'v3',
   description='A python package implementing various econometrical tests and estimators',
   packages = find_packages(),
   long_description=long_desc,