ComputationalStats

from statslibx import ComputationalStats
from statslibx.datasets import load_iris

data = load_iris()
cs = ComputationalStats(data, seed=42)

# Regresión lineal simple
modelo = cs.regression(X='sepal_length', y='petal_length')
print(modelo)
# 

# Ver resumen completo
s = modelo.summary()
print(s['metrics'])      # R², RMSE, MAE, AIC, BIC...
print(s['coefficients']) # tabla con t-stats y p-values
print(s['formula']['latex'])  # expresión LaTeX

# Predicciones
preds = modelo.predict([5.0, 6.0, 7.0])

# Regresión polinomial grado 3
modelo3 = cs.regression('sepal_length', 'petal_length', degree=3)
print(f"R² = {modelo3.r2:.4f}")

# Regresión múltiple
modelo_m = cs.regression(['sepal_length', 'sepal_width'], 'petal_length')
print(modelo_m.get_formula())

modelo = cs.linear_regression('sepal_length', 'petal_length')
modelo.plot()                   # 4 paneles: fit, residuales, Q-Q, histograma
modelo.plot(interactive=True)  # Plotly interactivo
modelo.plot(plot_type='qq')    # Solo Q-Q plot

modelo = cs.polynomial_regression('sepal_length', 'petal_length', degree=3)
print(f"Fórmula: {modelo.get_formula()}")
print(f"LaTeX:   {modelo.latex_expr}")

import numpy as np
preds = modelo.predict(np.linspace(4.5, 8.0, 20))

resultado = cs.find_best_degree('sepal_length', 'petal_length', max_degree=5, metric='bic')
print(f"Mejor grado: {resultado['degree']}")
print(f"R²:  {resultado['r2']:.4f}")
print(f"BIC: {resultado['bic']:.4f}")

# Tabla comparativa
for r in resultado['all_results']:
    print(f"Grado {r['degree']}: R²={r['r2']:.4f}  AIC={r['aic']:.2f}  BIC={r['bic']:.2f}")

# Graficar el mejor modelo
resultado['model'].plot()

points = [(0, 1), (1, 3), (2, 2), (3, 5), (4, 4)]

# Lagrange
interp = cs.interpolation(points, method='lagrange')
print(interp.summary())
print(f"LaTeX: {interp.latex_expr}")
interp.plot()
interp.plot(interactive=True)

# Newton
interp_n = cs.interpolation(points, method='newton')
print(f"En x=2.5: {interp_n.predict(2.5):.4f}")

# Evaluar en múltiples puntos
import numpy as np
x_new = np.linspace(0, 4, 100)
y_new = interp.predict(x_new)

# Spline cúbico
interp_s = cs.interpolation(points, method='spline', spline_degree=3)

# Bootstrap de la media — IC 95%
boot = cs.bootstrapping('sepal_length', n_samples=2000, statistic='mean')
s = boot.summary()
print(f"Estadístico original: {s['original_statistic']:.4f}")
print(f"Sesgo:               {s['bias']:.6f}")
print(f"Error estándar:      {s['std_error']:.4f}")

ic = s['confidence_interval_95%']
print(f"IC Percentil: {ic['percentile']}")
print(f"IC Básico:    {ic['basic']}")
print(f"IC Normal:    {ic['normal']}")

boot.plot()
boot.plot(interactive=True)

# Bootstrap de la mediana con IC 99%
boot_med = cs.bootstrapping('petal_length', statistic='median', confidence_level=0.99)

# Bootstrap con función personalizada (coeficiente de variación)
boot_cv = cs.bootstrapping(
    'sepal_width',
    statistic='custom',
    custom_func=lambda x: x.std() / x.mean()
)

resultado = cs.k_means(k=3, init_method='kmeans++')
print(f"Inercia:     {resultado['inertia']:.2f}")
print(f"Silhouette:  {resultado['silhouette_score']:.4f}")
print(f"Iteraciones: {resultado['n_iterations']}")

# Añadir etiquetas al DataFrame
data['cluster'] = resultado['labels']
print(data.groupby('cluster').mean())

elbow = cs.elbow_method(max_k=8)
# {'k_values': [2,3,...,8], 'inertias': [...], 'silhouettes': [...]}

import matplotlib.pyplot as plt
fig, axes = plt.subplots(1, 2, figsize=(12, 5))

axes[0].plot(elbow['k_values'], elbow['inertias'], 'bo-')
axes[0].set(xlabel='k', ylabel='Inercia', title='Método del Codo')
axes[0].grid(alpha=0.3)

axes[1].plot(elbow['k_values'], elbow['silhouettes'], 'rs-')
axes[1].set(xlabel='k', ylabel='Silhouette Score')
axes[1].grid(alpha=0.3)

plt.tight_layout(); plt.show()

resultado = cs.correlation_analysis(method='pearson')
print(resultado['correlation_matrix'])
print(resultado['p_values'])   # Solo para Pearson

resultado_sp = cs.correlation_analysis(method='spearman')

cs.plot_correlation_heatmap(method='pearson', annot=True)
cs.plot_correlation_heatmap(method='spearman', interactive=True)

print(cs.descriptive_statistics())
print(cs.descriptive_statistics(by='species'))  # Agrupado por categoría

cs.plot_distribution('sepal_length', kind='hist')
cs.plot_distribution('petal_length', by='species', kind='violin')
cs.plot_distribution('sepal_width', kind='box', interactive=True)

from statslibx import ComputationalStats
from statslibx.datasets import load_iris

data = load_iris()
cs = ComputationalStats(data, seed=42)

# 1. Mejor grado de regresión por BIC
best = cs.find_best_degree('sepal_length', 'petal_length', max_degree=5, metric='bic')
print(f"Mejor grado: {best['degree']}, R²={best['r2']:.4f}")
best['model'].plot()

# 2. Bootstrap de la media con IC 99%
boot = cs.bootstrapping('sepal_length', n_samples=5000, confidence_level=0.99)
s = boot.summary()
print(f"IC 99% percentil: {s['confidence_interval_99%']['percentile']}")
boot.plot(interactive=True)

# 3. Interpolación de Newton
points = [(4.5, 1.3), (5.0, 1.8), (5.5, 2.5), (6.0, 3.8), (7.0, 5.1)]
interp = cs.interpolation(points, method='newton')
print(f"LaTeX: {interp.latex_expr}")
interp.plot()

# 4. Método del codo + clustering óptimo
elbow = cs.elbow_method(max_k=6)
clusters = cs.k_means(k=3)
data['cluster'] = clusters['labels']
print(f"Silhouette: {clusters['silhouette_score']:.4f}")

# 5. Heatmap de correlación interactivo
cs.plot_correlation_heatmap(method='spearman', interactive=True)