[Statistics][Foreign Exchange]水曜日のドル円は上昇する、あるいは7月のドル円は下落する

import pandas as pd
import numpy as np
import scipy
from scipy import stats as st

from matplotlib import pylab as plt
import seaborn as sns
sns.set()
import statsmodels.api as sm
from statsmodels.stats.multicomp import pairwise_tukeyhsd

import datetime
import time

USDJPY_DAY = pd.read_csv("/Users/Documents/USDJPY_DAY.csv")
USDJPY_DAY["date"] = pd.to_datetime(USDJPY_DAY["date"])

# pandasの計算は遅いので、全てnumpy配列に変換
date = np.array(USDJPY_DAY["date"])
opening = np.array(USDJPY_DAY["opening"])#始値
high = np.array(USDJPY_DAY["high"])#高値
low = np.array(USDJPY_DAY["low"])#低値
closing = np.array(USDJPY_DAY["closing"])#終値

dict_week = {"Monday":0, "Tuesday":0, "Wednesday":0, "Thursday":0, "Friday":0}# 値上がりする日の集計
dict_all = {"Monday":0, "Tuesday":0, "Wednesday":0, "Thursday":0, "Friday":0}# 曜日ごとの日数

for i in range(len(date)):# dict_weekもdict_allも一度に計算
    temp = datetime.datetime.strptime(date[i], "%Y-%m-%d")
    temp = temp.strftime("%A")
    dict_all += 1
    if closing[i] >= opening[i]:
        dict_week += 1

# 割合に変換
ratio_of_increase = [a/b for a, b in zip(dict_week.values(), dict_all.values())]

# %%まとめて選択
plt.bar(dict_week.keys(), ratio_of_increase, color = "skyblue", alpha = 0.7)
plt.title("USDJPY_DAY")
plt.xlabel("week")
plt.ylabel("increase")
plt.savefig("/Users/Documents/week.png", format = "png", dpi = 300)

# 実際のデータ
observed_up = [413, 417, 419, 404, 391]

# 各曜日の日数から値上がりの回数を引いて、値下がりの回数を計算
observed_down = [total - up for total, up in zip([793, 795, 795, 795, 793], observed_up)]

# 2x5のコンティンジェンシーテーブルを作成
observed = [observed_up, observed_down]

# カイ二乗検定を実行
st.chi2_contingency(observed)

(2.532379386488967,
 0.6388472993439482,
 4,
 array([[408.18232183, 409.21178544, 409.21178544, 409.21178544,
         408.18232183],
        [384.81767817, 385.78821456, 385.78821456, 385.78821456,
         384.81767817]]))

USDJPY_DAY = pd.read_csv("/Users/Documents/USDJPY_DAY.csv")
USDJPY_DAY["date"] = pd.to_datetime(USDJPY_DAY["date"])
USDJPY_DAY.set_index("date", inplace = True)

# 曜日ごとにリサンプリング
USDJPY_DAY = USDJPY_DAY.resample("MS").agg({"opening": "first", "high": max, "low": min, "closing": "last"})

# pandasの計算は遅いので、全てnumpy配列に変換
opening = np.array(USDJPY_DAY["opening"])# 始値
high = np.array(USDJPY_DAY["high"])# 高値
low = np.array(USDJPY_DAY["low"])# 低値
closing = np.array(USDJPY_DAY["closing"])# 終値

USDJPY_DAY["positive"] = USDJPY_DAY["closing"] - USDJPY_DAY["opening"] > 0
# 月ごとにpositiveの平均を計算
monthly_positive_ratio = USDJPY_DAY.groupby(USDJPY_DAY.index.month)["positive"].mean()

# χ二乗検定を行う。
# 各月のpositiveの回数を計算
positive_counts = USDJPY_DAY.groupby(USDJPY_DAY.index.month)['positive'].sum()

# 各月のそれ以外の回数を計算
negative_counts = USDJPY_DAY.groupby(USDJPY_DAY.index.month).size() - positive_counts

# 2x12のコンティンジェンシーテーブルを作成
observed = pd.concat([positive_counts, negative_counts], axis=1)

# カイ二乗検定を実行
st.chi2_contingency(observed)

(8.411507032102229,
 0.6760359835065202,
 11,
 array([[7.54054054, 7.45945946],
        [7.54054054, 7.45945946],
        [8.04324324, 7.95675676],
        [8.04324324, 7.95675676],
        [8.04324324, 7.95675676],
        [8.04324324, 7.95675676],
        [8.04324324, 7.95675676],
        [7.54054054, 7.45945946],
        [7.54054054, 7.45945946],
        [7.54054054, 7.45945946],
        [7.54054054, 7.45945946],
        [7.54054054, 7.45945946]]))

# グラフを描画
plt.figure(figsize=(10, 6))
plt.bar(monthly_positive_ratio.index, monthly_positive_ratio.values, color = "skyblue", alpha = 0.7)
plt.xlabel('Month')
plt.ylabel('Positive Ratio')
plt.title('Monthly Positive Ratio of Closing - Opening')
plt.xticks(range(1, 13), ['Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec'])
plt.show()

# 7月
# 7月のpositiveの回数を計算
july_positive_count = USDJPY_DAY[USDJPY_DAY.index.month == 7]['positive'].sum()

# 7月の日数を計算
july_days = len(USDJPY_DAY[USDJPY_DAY.index.month == 7])

# 他の月のpositiveの割合を計算
other_months_positive_ratio = USDJPY_DAY[USDJPY_DAY.index.month != 7]['positive'].mean()

# 二項検定を実行
p = st.binom_test(july_positive_count, n=july_days, p=other_months_positive_ratio)

print(f"P-value = {p}")
P-value = 0.13208599522026054