patchv1

.gitignore

@@ -0,0 +1,2 @@
+old/**
+*zip

load_dataset.py

@@ -9,6 +9,92 @@ def safe_float(x):
     return np.nan
 
 
+
+def analisis_univariado(dfi, target=None, continuas=[], discretas=[]):
+    if target is None:
+        raise ValueError("No target variable provided")
+
+    import pandas as pd
+    from scipy.stats import shapiro, ttest_ind, mannwhitneyu, chi2_contingency, fisher_exact
+
+#    label_columns = ['sexo', 'hist fam', 'edad diag', 'IMC', 'glu ayu', 'glu 120','A1c']
+    label_columns = dfi.drop(target, axis=1).columns
+
+    # Separar el target en dos grupos: N positivo y N negativo
+    groups = dfi[target].unique()
+    if len(groups) != 2:
+        raise ValueError("Target variable must have exactly two unique values")
+
+    group1, group2 = groups
+    data_group1 = dfi[dfi[target] == group1][label_columns]
+    data_group2 = dfi[dfi[target] == group2][label_columns]
+
+    results = []
+
+    # Análisis de variables continuas
+    for var in continuas:
+        Xvar = dfi[var]
+        group1_values = data_group1[var]
+        group2_values = data_group2[var]
+
+        # Test de normalidad (Shapiro-Wilk)
+        stat, p = shapiro(Xvar)
+        normal = p >= 0.05
+
+        if normal:
+            # Distribución normal: media, desviación estándar, y test t
+            mean1, std1 = group1_values.mean(), group1_values.std()
+            mean2, std2 = group2_values.mean(), group2_values.std()
+            t_stat, t_pval = ttest_ind(group1_values, group2_values, equal_var=False)
+            results.append([
+                var, "Continua", "Normal",
+                f"Media: {mean1:.2f} (Grupo 1), {mean2:.2f} (Grupo 2)",
+                f"Desviación Est.: {std1:.2f} (Grupo 1), {std2:.2f} (Grupo 2)",
+                f"Test t: p={t_pval:.3f}"
+            ])
+        else:
+            # Distribución no normal: mediana, rango intercuartil, y test Mann-Whitney
+            median1, iqr1 = group1_values.median(), group1_values.quantile(0.75) - group1_values.quantile(0.25)
+            median2, iqr2 = group2_values.median(), group2_values.quantile(0.75) - group2_values.quantile(0.25)
+            mw_stat, mw_pval = mannwhitneyu(group1_values, group2_values)
+            results.append([
+                var, "Continua", "No Normal",
+                f"Mediana: {median1:.2f} (Grupo 1), {median2:.2f} (Grupo 2)",
+                f"RIC: {iqr1:.2f} (Grupo 1), {iqr2:.2f} (Grupo 2)",
+                f"Mann-Whitney: p={mw_pval:.3f}"
+            ])
+
+    # Análisis de variables discretas
+    for var in discretas:
+        freq_table = dfi.groupby([target, var]).size().unstack(fill_value=0)
+        percentages = freq_table.div(freq_table.sum(axis=1), axis=0) * 100
+
+        # Pruebas estadísticas
+        if freq_table.shape[1] == 2:
+            # Test exacto de Fisher
+            _, fisher_pval = fisher_exact(freq_table.values)
+            test_result = f"Fisher Exact: p={fisher_pval:.3f}"
+        else:
+            # Test Chi cuadrado
+            chi2_stat, chi2_pval, _, _ = chi2_contingency(freq_table)
+            test_result = f"Chi2: p={chi2_pval:.3f}"
+
+        results.append([
+            var, "Discreta", "N/A",
+            f"Frecuencias: {freq_table.to_dict()}",
+            f"Porcentajes: {percentages.to_dict()}",
+            test_result
+        ])
+
+    # Crear DataFrame con los resultados
+    results_df = pd.DataFrame(results, columns=[
+        "Variable", "Tipo", "Distribución", "Medidas descriptivas", "Estadísticas", "Resultados Prueba"
+    ])
+    return results_df
+
+
+
+
 def load_data(Reload=False):
     if os.path.isfile('MODY_data.xlsx'):
         import pandas as pd

train.py

@@ -1,4 +1,4 @@
-from load_dataset import load_data
+from load_dataset import load_data, analisis_univariado
 from trainer import BinaryTuner
 import pandas as pd
 import numpy as np
@@ -7,13 +7,18 @@ warnings.filterwarnings("ignore")
 
 _, dms2, dms3, _ = load_data()
 
-mody2 = BinaryTuner(dms2, 'MODY2_label', seeds=[231964], test_size=0.2)
-mody2.fit()
-mody2.explain_model('GaussianNB', 'fulldataset-oversampled-mice', 231964)
-mody2.wrap_and_save()
+
+resultados = analisis_univariado(dms2, target="MODY2_label", continuas=['edad diag', 'IMC', 'glu ayu', 'glu 120','A1c'], discretas=['sexo', 'hist fam'])
+print(resultados)
 
 
-mody3 = BinaryTuner(dms3, 'MODY3_label', seeds=[536202], test_size=0.2)
-mody3.fit()
-mody3.explain_model('RandomForestClassifier', 'fulldataset-original-mice', 536202)
-mody3.wrap_and_save()
+# mody2 = BinaryTuner(dms2, 'MODY2_label', seeds=[231964], test_size=0.2)
+# mody2.fit()
+# mody2.explain_model('GaussianNB', 'fulldataset-oversampled-mice', 231964)
+# mody2.wrap_and_save()
+#
+#
+# mody3 = BinaryTuner(dms3, 'MODY3_label', seeds=[536202], test_size=0.2)
+# mody3.fit()
+# mody3.explain_model('RandomForestClassifier', 'fulldataset-original-mice', 536202)
+# mody3.wrap_and_save()