Assignment/did 157104

assignments/did/code/did-157104.py

@@ -0,0 +1,94 @@
+import os
+import pandas as pd
+import numpy as np
+from linearmodels.panel import PanelOLS
+
+# Cargar datos
+data_path = os.path.join("..", "data", "callaway-santanna.csv")
+df = pd.read_csv(data_path)
+
+# Renombrar columnas
+df.rename(columns={
+    "year": "t",
+    "countyreal": "i",
+    "first.treat": "treat_start"
+}, inplace=True)
+
+#  Variable de tiempo "k"
+df.loc[df['treat_start'] == 0, 'treat_start'] = np.nan
+df["k"] = df["t"] - df["treat_start"]
+
+# Event dummies w/Nans
+k_dummies = pd.get_dummies(df["k"], prefix="k", dummy_na=True, dtype=int)
+df = pd.concat([df, k_dummies], axis=1)
+
+# Indices
+df = df.set_index(["i", "t"])
+
+# ATT
+mask_treated = df['treat_start'].notna()
+mask_full_panel = ~df['lemp'].isna().groupby(level='i').any()
+valid_units = mask_full_panel[mask_full_panel].index
+mask = mask_treated & df.index.get_level_values('i').isin(valid_units)
+df_att = df[mask].copy()
+all_k_dummies = [col for col in df_att.columns if col.startswith('k_')]
+
+# Periodo referencia
+k_dummies = [k for k in all_k_dummies if k not in ['k_-1.0', 'k_nan']]
+
+# Dummies
+k_dummies_final = [k for k in k_dummies if df_att[k].nunique() > 1]
+indep = df_att[k_dummies_final]
+
+spec0 = PanelOLS(
+    df_att["lemp"],
+    indep,
+    entity_effects=True,
+    time_effects=True,
+    drop_absorbed=True
+)
+
+res0 = spec0.fit(cov_type="clustered")
+
+restriction0 = np.array([
+    [1, 0, 0, 0, 0, 0],  
+    [0, 1, 0, 0, 0, 0],  
+    [0, 0, 1, 0, 0, 0]   
+])
+values0 = np.array([0, 0, 0])  
+f0=res0.wald_test(restriction0, values0)
+
+anticipation0 = False
+att0 = '-'
+
+# ATT - All units
+vars1 = [col for col in df.columns if col.startswith('k_') and col != 'k_-1.0']
+# Definir dep1 e indep 
+dep1 = df["lemp"]
+indep1 = df[vars1]
+
+# 
+m1 = PanelOLS(
+    dependent=dep1,
+    exog=indep1,
+    entity_effects=True,
+    time_effects=True,
+    drop_absorbed=True
+)
+
+# Estimar modelo
+res1 = m1.fit(cov_type="clustered")
+
+# Probar efectos anticipados
+restriction1 = np.array([
+    [1, 0, 0, 0, 0, 0, 0],  
+    [0, 1, 0, 0, 0, 0, 0],  
+    [0, 0, 1, 0, 0, 0, 0]   
+])
+values1 = np.array([0, 0, 0])
+f1 = res1.wald_test(restriction1, values1)
+
+# Anticipation 1 effect
+anticipation1 = False
+# Signo git aAt1
+att1 = '-'

assignments/ivs/code/ivs-157104.py

@@ -0,0 +1,91 @@
+# Imports
+import pandas as pd
+import itertools
+import statsmodels.api as sm
+from linearmodels.iv import IV2SLS
+
+# Read data
+PATH = os.path.join('..', 'data', 'raw.csv')
+df = pd.read_csv(PATH)
+df = df[df['yob'] > 1939]
+
+# Crear dummies para 'qob'
+dummies_qob = pd.get_dummies(df['qob'], prefix='qob', dtype=int)
+# Crear dummies para 'yob'
+dummies_yob = pd.get_dummies(df['yob'], prefix='yob', dtype=int)
+# Agregar las dummies al DataFrame original
+df = pd.concat([df, dummies_yob, dummies_qob], axis=1)
+
+# Obtener listas de los años y trimestres
+years = range(1940, 1950)  # Años de 1940 a 1949
+quarters = range(1, 5)     # Trimestres del 1 al 4
+combinations = itertools.product(years, quarters)
+
+# Iterar y crear las columnas de interacción
+for year, quarter in combinations:
+    interaction_col_name = f'yob_{year}_qob_{quarter}'
+
+    # Verificar que ambas columnas existen antes de realizar la multiplicación
+    if f'yob_{year}' in df.columns and f'qob_{quarter}' in df.columns:
+        df[interaction_col_name] = df[f'yob_{year}'] * df[f'qob_{quarter}']
+    else:
+        print(f"Columnas faltantes: yob_{year} o qob_{quarter}")
+
+# Definir la variable dependiente
+y = df['lwklywge']
+
+# Crear las dummies para 'yob'
+yob_dummies = pd.get_dummies(df['yob'], prefix='yob', dtype=int)
+yob_dummies = yob_dummies.drop(columns=['yob_1949'], errors='ignore')
+
+# Definir las variables independientes (constante y controles)
+X = df[['race', 'married', 'smsa', 'neweng', 'midatl', 'enocent', 'wnocent', 
+        'soatl', 'esocent', 'wsocent', 'mt', 'educ']]
+
+# Combinar los controles con las dummies del año de nacimiento
+X = pd.concat([X, yob_dummies], axis=1)
+
+# Agregar la constante al DataFrame de variables independientes
+X = sm.add_constant(X)
+
+# Ajustar el modelo ingenuo usando OLS (Regresión Lineal Ordinaria)
+model = sm.OLS(y, X).fit(cov_type='HC3')  # HC3 para errores estándar robustos
+
+# Almacenar los resultados en 'res0'
+res0 = model.summary()
+
+# Crear la lista de interacciones como instrumentos, omitiendo el último trimestre de cada año
+instrument_cols = []
+for year in years:
+    for quarter in quarters:
+        if not (quarter == 4):  # Omitir el último trimestre como referencia
+            instrument_cols.append(f'yob_{year}_qob_{quarter}')
+
+# Crear la matriz de instrumentos
+instruments = df[instrument_cols]
+
+# Verificar que las columnas de instrumentos existen
+print(f"Instrumentos seleccionados: {instruments.columns.tolist()}")
+
+# Redefinir las variables exógenas excluyendo 'educ' (que es endógena)
+X_iv = df[['race', 'married', 'smsa', 'neweng', 'midatl', 'enocent', 'wnocent',
+           'soatl', 'esocent', 'wsocent', 'mt']]
+
+# Agregar las dummies del año de nacimiento, excluyendo 'yob_1949' como referencia
+X_iv = pd.concat([X_iv, yob_dummies], axis=1)
+
+# Agregar la constante
+X_iv = sm.add_constant(X_iv)
+
+# Definir la variable endógena (educ)
+endog = df['educ']
+
+# Ajustar el modelo 2SLS
+iv_model = IV2SLS(dependent=y, exog=X_iv, endog=endog, instruments=instruments).fit()
+
+# Almacenar los resultados en 'res1'
+res1 = iv_model.summary
+
+bias = True
+
+bias_sign = '+'

assignments/rct/code/rct-157104.py

@@ -0,0 +1,48 @@
+# Imports
+import os
+import pandas as pd
+import statsmodels.api as sm
+
+# Load data
+PATH = os.path.join('..', 'data', 'raw.csv')
+df = pd.read_csv(PATH)
+
+    # Rename columns
+df.columns = ['id', 'dark', 'views', 'time', 'purchase', 'mobile', 'location']
+
+# Map columns to numeric types
+df.replace(
+    to_replace={
+        'dark':{'A':'0', 'B':'1'},
+        'mobile':{'Mobile':'1', 'Desktop':'0'},
+        'purchase':{'No':'0', 'Yes':'1'},
+        'location':{'Northern Ireland':'Ireland'}
+    },
+    inplace=True
+)
+
+# Convert strings -> ints
+df[['dark', 'mobile', 'purchase']] = df[['dark', 'mobile', 'purchase']].astype(int)
+# A minusculas
+df['location'] = df['location'].str.lower()
+
+# One-hot encoding
+df = pd.get_dummies(
+    data = df,
+    prefix = '',
+    prefix_sep = '',
+    columns = ['location'],
+    dtype = int
+)
+
+# Constant
+df['const'] = 1
+
+# Declare linear model
+spec = sm.OLS(
+    endog = df['purchase'],
+    exog = df[['const', 'ireland','scotland','wales','dark']],
+    hasconst= True
+)
+
+model = spec.fit()

assignments/rdd/code/rdd.ipynb

@@ -11,7 +11,7 @@
     },
     {
       "cell_type": "code",
-      "execution_count": null,
+      "execution_count": 1,
       "metadata": {
         "id": "HWkWFR_iMk8h"
       },