diffindiff 2.0.3__tar.gz → 2.0.5__tar.gz

{diffindiff-2.0.3 → diffindiff-2.0.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: diffindiff
-Version: 2.0.3
+Version: 2.0.5
 Summary: diffindiff: Python library for convenient Difference-in-Differences Analyses
 Author: Thomas Wieland
 Author-email: geowieland@googlemail.com
@@ -45,7 +45,7 @@ Thomas Wieland [ORCID](https://orcid.org/0000-0001-5168-9846) [EMail](mailto:geo
     - Group- or individual-specific treatment effects
     - Group- or individual-specific time trends
     - Including covariates
-    - Including fter-treatment period
+    - Including after-treatment period
     - Triple Difference (DDD)
     - Own counterfactuals
     - Bonferroni correction for treatment effects

{diffindiff-2.0.3 → diffindiff-2.0.5}/README.md

@@ -23,7 +23,7 @@ Thomas Wieland [ORCID](https://orcid.org/0000-0001-5168-9846) [EMail](mailto:geo
     - Group- or individual-specific treatment effects
     - Group- or individual-specific time trends
     - Including covariates
-    - Including fter-treatment period
+    - Including after-treatment period
     - Triple Difference (DDD)
     - Own counterfactuals
     - Bonferroni correction for treatment effects

{diffindiff-2.0.3 → diffindiff-2.0.5}/diffindiff/didanalysis.py

@@ -4,8 +4,8 @@
 # Author:      Thomas Wieland 
 #              ORCID: 0000-0001-5168-9846
 #              mail: geowieland@googlemail.com              
-# Version:     2.0.3
-# Last update: 2025-04-18 10:24
+# Version:     2.0.5
+# Last update: 2025-04-19 10:23
 # Copyright (c) 2025 Thomas Wieland
 #-----------------------------------------------------------------------
 
@@ -15,6 +15,7 @@ from statsmodels.formula.api import ols
 import numpy as np
 import matplotlib.pyplot as plt
 from matplotlib.dates import DateFormatter
+import json
 import diffindiff.didtools
 
 
@@ -27,8 +28,7 @@ class DiffModel:
         did_modeldata,
         did_modelpredictions,
         did_model_statistics,
-        did_olsmodel,
-        did_prediction_intervals
+        did_olsmodel
         ):
 
         self.data = [
@@ -37,9 +37,8 @@ class DiffModel:
             did_modeldata, 
             did_modelpredictions, 
             did_model_statistics, 
-            did_olsmodel,
-            did_prediction_intervals
-            ]
+            did_olsmodel
+            ]    
 
     def treatment_statistics(
         self,
@@ -165,11 +164,11 @@ class DiffModel:
             treatment_diagnostics_rows.append({
                 "Treatment": value["treatment"],
                 "Type of adoption": adoption_type,
-                "No-treatment control group": no_treatment,
-                "Treatment group (N)": treatment_group_size,
-                "Control group (N)": control_group_size,
+                "No-treatment control group": no_treatment,                
                 "Parallel trends (pre)": is_parallel,
-                "Format": value["treatment_format"]
+                "Format": value["treatment_format"],
+                "Treatment group (N)": treatment_group_size,
+                "Control group (N)": control_group_size
             })
 
             if no_treatment == "NO" and adoption_type == "Simultaneous":
@@ -230,8 +229,8 @@ class DiffModel:
                     })
                 average_after_treatment_effects = pd.DataFrame(average_after_treatment_effects_rows)
                 if len (average_after_treatment_effects) == 1:
-                    average_after_treatment_effects.at[0, ""] = "Average after treatment effect"
-                treatment_effects_df = pd.concat([treatment_effects_df, average_after_treatment_effects], ignore_index=True)                  
+                    average_after_treatment_effects.at[0, ""] = "Average after-treatment effect"
+                treatment_effects_df = pd.concat([treatment_effects_df, average_after_treatment_effects], ignore_index=True)
 
             if ("control_group_baseline" in model_results and not model_config["FE_unit"]) or ("control_group_baseline" in model_results and baseline_components):            
                 control_group_baseline = model_results["control_group_baseline"]                
@@ -287,7 +286,25 @@ class DiffModel:
                 if len (non_treatment_time_effect) == 1:
                     non_treatment_time_effect.at[0, ""] = "Non-treatment time effect"
                 treatment_effects_df = pd.concat([treatment_effects_df, non_treatment_time_effect], ignore_index=True)
-                                
+
+            if "after_treatment_time_effects" in model_results:          
+                after_treatment_time_effects = model_results["after_treatment_time_effects"]            
+                after_treatment_time_effects_rows = []
+                for key, value in after_treatment_time_effects.items():
+                    after_treatment_time_effects_rows.append({
+                        "": "After-treatment time effect " + value["Coefficient"],
+                        "Estimate": value["Estimate"],
+                        "SE": value["SE"],
+                        "t": value["t"],
+                        "p": value["p"],
+                        "CI lower": value["CI_lower"],
+                        "CI upper": value["CI_upper"]
+                    })
+                after_treatment_time_effects = pd.DataFrame(after_treatment_time_effects_rows)
+                if len (after_treatment_time_effects) == 1:
+                    after_treatment_time_effects.at[0, ""] = "After-treatment time effect"
+                treatment_effects_df = pd.concat([treatment_effects_df, after_treatment_time_effects], ignore_index=True)
+
         if "individual_treatment_effects" in model_results:
 
             individual_treatment_effects = model_results["individual_treatment_effects"]
@@ -535,7 +552,6 @@ class DiffModel:
             else:
                 print ("Covariates                 NO")
 
-        print("")
         print("Fixed effects")        
         if model_config["FE_unit"]:
             print (" Units                     YES")
@@ -570,9 +586,21 @@ class DiffModel:
             index = treatment_diagnostics_df.columns)
         treatment_diagnostics_df_t = treatment_diagnostics_df_t.iloc[1:]
         print(treatment_diagnostics_df_t)
+        if model_config["no_treatments"] > 1:
+            untreated = diffindiff.didtools.untreated_units(
+                data = model_data,
+                unit_col = model_config["unit_col"],
+                treatment_col = model_config["treatment_col"]
+                )
+            print ("Units with >=1 treatment(s): " + str(untreated[0]) + ", non-treated units: " + str(untreated[1]))
+        if len(no_control_conditions) > 0:
+            if len(no_control_conditions) == 1:
+                print("NOTE: Treatment " + no_control_conditions[0], " has no control conditions")
+            else:
+                print("NOTE: Treatments " + ", ".join(no_control_conditions), "have no control conditions")  
 
         print("-" * total_width)
-        print ("Input data diagnostixx") # TODO ?? AENDERN
+        print ("Input data diagnostics")
         if modeldata_isbalanced:
             print ("Balanced panel data        YES")
         else:
@@ -591,12 +619,6 @@ class DiffModel:
         print ("Adj. R-Squared             " + str(round(model_statistics["rsquared_adj"], 3)))
         print ("===============================================================")
 
-        if len(no_control_conditions) > 0:
-            if len(no_control_conditions) == 1:
-                print("NOTE: Treatment " + ",".join(no_control_conditions), "has no control conditions")
-            else:
-                print("NOTE: Treatments " + ", ".join(no_control_conditions), "have no control conditions")
-
         return self
 
     def plot_treatment_effects(
@@ -762,9 +784,16 @@ class DiffModel:
         ols_model = self.data[5]
         return ols_model
     
-    def prediction_intervals(self):
+    def prediction_intervals(
+        self,
+        confint_alpha = 0.05
+        ):
+
+        ols_model = self.data[5]
+
+        prediction_intervals = ols_model.get_prediction()
+        prediction_intervals = prediction_intervals.summary_frame(alpha = confint_alpha)
 
-        prediction_intervals = self.data[6]
         return prediction_intervals
     
     def placebo(
@@ -807,9 +836,13 @@ class DiffModel:
         TT_col_ = "TT_" + treatment
         TGxTT_ = "Placebo_" + treatment
         if TG_col is None and TG_col_ not in model_config["TG_col"]:
-            raise ValueError("Cannot find treatment group identification variable for treatment " + treatment + ". Please state TG_col = [treatment_group_dummy].")
+            raise ValueError("No treatment group identification variable for treatment " + treatment + ". Please state TG_col = [treatment_group_dummy].")
         if TT_col is None and TT_col_ not in model_config["TT_col"]:
-            raise ValueError("Cannot findt treatment time variable for treatment " + treatment + ". Please state TG_col = [treatment_time_dummy].")
+            raise ValueError("No treatment time variable for treatment " + treatment + ". Please state TG_col = [treatment_time_dummy].")
+        if TG_col is not None:
+            TG_col_ = TG_col
+        if TT_col is not None:
+            TT_col_ = TT_col
 
         unit_col = model_config["unit_col"]
         time_col = model_config["time_col"]
@@ -1270,6 +1303,7 @@ def did_analysis(
     TG_col: list = [],
     TT_col: list = [],
     after_treatment_col: list = [],
+    ATT_col: list = [],
     pre_post: bool = False,
     log_outcome: bool = False,
     log_outcome_add = 0.01,
@@ -1407,6 +1441,16 @@ def did_analysis(
             )        
         cols_relevant = cols_relevant + after_treatment_col
 
+    if ATT_col is not None or (isinstance (ATT_col, list) and len(ATT_col) > 0):
+        if isinstance (ATT_col, str):
+            ATT_col = [ATT_col]
+        ATT_col = [entry for entry in ATT_col if entry is not None]
+        diffindiff.didtools.check_columns(
+            df = data,
+            columns = ATT_col
+            )        
+        cols_relevant = cols_relevant + ATT_col
+
     if TG_col is not None or (isinstance (TG_col, list) and len(TG_col) > 0):
         if isinstance (TG_col, str):
             TG_col = [TG_col]
@@ -1504,6 +1548,8 @@ def did_analysis(
 
     if len(after_treatment_col) > 0:
         did_formula = did_formula + f' + {" + ".join(after_treatment_col)}'
+    if len(ATT_col) > 0:
+        did_formula = did_formula + f' + {" + ".join(ATT_col)}'
     
     if FE_unit:
         unit_col_todummies = diffindiff.didtools.to_dummies(
@@ -1652,6 +1698,7 @@ def did_analysis(
         "freq": freq,
         "date_format": date_format,
         "after_treatment_col": after_treatment_col,
+        "ATT_col": ATT_col,
         "pre_post": pre_post,
         "FE_unit": FE_unit,
         "FE_time": FE_time,
@@ -1759,6 +1806,20 @@ def did_analysis(
                 "CI_upper": float(coef_conf_intervals.loc[after_treatment, 1]),
                 }
         model_results["average_after_treatment_effects"] = AATE
+    
+    if (any(col in ols_coefficients for col in ATT_col)):
+        ATT = {}        
+        for i, ATT_ in enumerate(ATT_col):
+            ATT[i] = {
+                "Coefficient": ATT_,
+                "Estimate": ols_coefficients[ATT_],
+                "SE": float(coef_standarderrors[ATT_]),
+                "t": float(coef_t[ATT_]),
+                "p": float(coef_p[ATT_]),
+                "CI_lower": float(coef_conf_intervals.loc[ATT_, 0]),
+                "CI_upper": float(coef_conf_intervals.loc[ATT_, 1]),
+                }            
+        model_results["after_treatment_time_effects"] = ATT
 
     if DDD:
 
@@ -1990,9 +2051,6 @@ def did_analysis(
 
     model_predictions = ols_model.predict()
     
-    prediction_intervals = ols_model.get_prediction()
-    prediction_intervals = prediction_intervals.summary_frame(alpha = confint_alpha)
-    
     model_statistics = {
         "rsquared": ols_model.rsquared,
         "rsquared_adj": ols_model.rsquared_adj,
@@ -2005,8 +2063,7 @@ def did_analysis(
         data,
         model_predictions,
         model_statistics,
-        ols_model,
-        prediction_intervals
+        ols_model
         ) 
 
     return did_model_output

{diffindiff-2.0.3 → diffindiff-2.0.5}/diffindiff/diddata.py

@@ -4,8 +4,8 @@
 # Author:      Thomas Wieland 
 #              ORCID: 0000-0001-5168-9846
 #              mail: geowieland@googlemail.com              
-# Version:     2.0.3
-# Last update: 2025-04-18 10:24
+# Version:     2.0.5
+# Last update: 2025-04-19 10:23
 # Copyright (c) 2025 Thomas Wieland
 #-----------------------------------------------------------------------
 
@@ -163,7 +163,7 @@ class DiffTreatment:
             else:
                 print (" Treatment Period:        " + str(value["treatment_period_start"]) + " - " + str(value["treatment_period_end"])+ (" (") + str(value["treatment_period"]) + " " + treatment_meta["frequency"] + ")")
             if treatment_config[key]["after_treatment_period"]:
-                print (" After treatment period:  " + str(value["treatment_period_end"]) + " - " + str(value["study_period_end"]) + " (" + str(value["after_treatment_period_N"]) + " " + treatment_meta["frequency"] + ")")
+                print (" After-treatment period:  " + str(value["treatment_period_end"]) + " - " + str(value["study_period_end"]) + " (" + str(value["after_treatment_period_N"]) + " " + treatment_meta["frequency"] + ")")
         
 def create_treatment (
     study_period,
@@ -722,12 +722,15 @@ class DiffData:
         TG_col = [None]*len(treatment_cols)
         treatment_col = [None]*len(treatment_cols)
         after_treatment_col = [None]*len(treatment_cols)
+        ATT_col = [None]*len(treatment_cols)
         for key, value in treatment_cols.items():
             TG_col[key] = groups_config[key]["TG_col"]
             TT_col[key] = value["TT_col"]
             treatment_col[key] = value["treatment_name"]
             if value["after_treatment_name"] is not None:
                 after_treatment_col[key] = value["after_treatment_name"]
+            if value["ATT_col"] is not None:
+                ATT_col[key] = value["ATT_col"]
         
         group_benefit = []
         if groups_config[0]["DDD"]:
@@ -742,6 +745,7 @@ class DiffData:
             time_col = "t",
             outcome_col = outcome_col_original,
             after_treatment_col = after_treatment_col,
+            ATT_col = ATT_col,
             pre_post = treatment_meta["pre_post"],
             log_outcome = log_outcome,
             FE_unit = FE_unit,

{diffindiff-2.0.3 → diffindiff-2.0.5}/diffindiff/didtools.py

@@ -4,8 +4,8 @@
 # Author:      Thomas Wieland 
 #              ORCID: 0000-0001-5168-9846
 #              mail: geowieland@googlemail.com              
-# Version:     2.0.3
-# Last update: 2025-04-18 12:08
+# Version:     2.0.5
+# Last update: 2025-04-19 10:23
 # Copyright (c) 2025 Thomas Wieland
 #-----------------------------------------------------------------------
 
@@ -186,11 +186,22 @@ def treatment_group_col(
 def untreated_units(
     data: pd.DataFrame,
     unit_col: str,
-    time_col: str,
-    treatment_col: str   
+    treatment_col: list   
     ):
-    # TODO ??
-    pass
+
+    unit_sum = data.groupby(unit_col)[treatment_col].sum().sum(axis=1).reset_index(name="sum")
+
+    units_treated = unit_sum.loc[unit_sum["sum"] > 0, unit_col]
+    units_nontreated = unit_sum.loc[unit_sum["sum"] == 0, unit_col]
+    no_units_treated = len(units_treated)
+    no_units_nontreated = len(units_nontreated)
+
+    return [
+        no_units_treated,
+        no_units_nontreated,
+        units_treated,
+        units_nontreated
+        ]
 
 def is_parallel(
     data: pd.DataFrame,

{diffindiff-2.0.3 → diffindiff-2.0.5}/diffindiff/tests/tests_diffindiff.py

@@ -4,8 +4,8 @@
 # Author:      Thomas Wieland 
 #              ORCID: 0000-0001-5168-9846
 #              mail: geowieland@googlemail.com              
-# Version:     2.0.3
-# Last update: 2025-04-18 10:24
+# Version:     2.0.5
+# Last update: 2025-04-19 10:23
 # Copyright (c) 2025 Thomas Wieland
 #-----------------------------------------------------------------------
 
@@ -13,7 +13,7 @@
 import pandas as pd
 from diffindiff.didanalysis import DiffModel, did_analysis
 from diffindiff.diddata import DiffGroups, create_groups, DiffTreatment, create_treatment, DiffData, merge_data, create_data
-from diffindiff.didtools import treatment_group_col
+
 
 # Example 1: Effect of a curfew in German counties in the first
 # wave of the COVID-19 pandemic (DiD pre-post analysis)

{diffindiff-2.0.3 → diffindiff-2.0.5}/diffindiff.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: diffindiff
-Version: 2.0.3
+Version: 2.0.5
 Summary: diffindiff: Python library for convenient Difference-in-Differences Analyses
 Author: Thomas Wieland
 Author-email: geowieland@googlemail.com
@@ -45,7 +45,7 @@ Thomas Wieland [ORCID](https://orcid.org/0000-0001-5168-9846) [EMail](mailto:geo
     - Group- or individual-specific treatment effects
     - Group- or individual-specific time trends
     - Including covariates
-    - Including fter-treatment period
+    - Including after-treatment period
     - Triple Difference (DDD)
     - Own counterfactuals
     - Bonferroni correction for treatment effects

{diffindiff-2.0.3 → diffindiff-2.0.5}/setup.py

@@ -7,7 +7,7 @@ def read_README():
     
 setup(
     name='diffindiff',
-    version='2.0.3',
+    version='2.0.5',
     description='diffindiff: Python library for convenient Difference-in-Differences Analyses',
     packages=find_packages(include=["diffindiff", "diffindiff.tests"]),
     include_package_data=True,