LoginSignup
1
1

More than 5 years have passed since last update.

StanとRでベイズ統計モデリング(アヒル本)をPythonにしてみる - 10.1 パラメータの識別可能性

Last updated at Posted at 2018-08-19

実行環境

インポート

import numpy as np
from scipy import stats
import pandas as pd
import statsmodels.api as sm
import pystan
import matplotlib.pyplot as plt
from matplotlib.figure import figaspect
%matplotlib inline

データ読み込み

category = pd.read_csv('./data/data-category.txt')
usagitokame = pd.read_csv('./data/data-usagitokame.txt')

10.1 パラメータの識別可能性

10.1.3 ラベルスイッチング

np.random.seed(123)

N = 200
a = 0.3
d1 = np.random.normal(loc=0, scale=2, size=N)
d2 = np.random.normal(loc=-4, scale=1, size=N)
d3 = np.random.uniform(size=N) <= a
Y = d1*d3 + d2*(1-d3)
X = np.linspace(Y.min(), Y.max(), 100)

plt.figure(figsize=figaspect(3/4))
ax = plt.axes()
ax.hist(Y, facecolor='w', edgecolor='k', density=True)
ax.plot(X, a*stats.norm.pdf(X, loc=0, scale=2), c='k', linestyle='dashed')
ax.plot(X, (1-a)*stats.norm.pdf(X, loc=-4, scale=1), c='k', linestyle='dashed')
ax.plot(X, a*stats.norm.pdf(X, loc=0, scale=2) + (1-a)*stats.norm.pdf(X, loc=-4, scale=1), c='k', alpha=0.4, linestyle='solid', linewidth=3)
plt.setp(ax, xlabel='y', ylabel='density')
plt.show()

fig10-1.png

10.1.4 多項ロジスティック回帰

X = sm.add_constant(category.iloc[:, :-1])
X['Age'] /= 100
X['Income'] /= 1000
data = dict(
    N=category.index.size,
    D=X.columns.size,
    K=category['Y'].max(),
    X=X,
    Y=category['Y']
)
fit = pystan.stan('./stan/model10-2.stan', data=data, seed=1234)

10.1.5 ウサギとカメ

np.random.seed(123)

G = 30
mu_pf = (0, 1.5)
pf = np.random.normal(loc=mu_pf, scale=1, size=(G, 2))
d = pd.DataFrame(np.argsort(pf, axis=1)+1, columns=('Loser', 'Winner'))
tbl = pd.crosstab([], d['Winner'])
tbl
data = dict(
    N=usagitokame.columns.size,
    G=usagitokame.index.size,
    LW=usagitokame
)
fit = pystan.stan('./stan/model10-3.stan', data=data, seed=1234)
1
1
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
1
1