5-14. ジェンダーギャップ指数を偏差値で再計算すると日本は何位なのか？

import numpy as np
import numpy.random as nr
import pandas as pd
from copy import deepcopy
import matplotlib.pyplot as plt

#データ読み込み
data = pd.read_csv('Global Gender Gap Subindex 2021.csv', encoding='utf-8', header=0)

#国名
data_country = np.array(data['Country Name'])
print('data_country[:5]   = {}'.format(data_country[:5]))

#4指標を抽出
data_health = np.array(data['Global Gender Gap Health and Survival Subindex'])
data_education = np.array(data['Global Gender Gap Educational Attainment Subindex'])
data_economy = np.array(data['Global Gender Gap Economic Participation and Opportunity Subindex'])
data_politics = np.array(data['Global Gender Gap Political Empowerment subindex'])
print('data_health[:5]    = {}'.format(data_health[:5]))
print('data_education[:5] = {}'.format(data_education[:5]))
print('data_economy[:5]   = {}'.format(data_economy[:5]))
print('data_politics[:5]  = {}'.format(data_politics[:5]))

#4指標を結合したもの
four = np.stack([data_health, data_education, data_economy, data_politics], axis=1)
#nanのある行を除く
country = data_country[~np.isnan(four).any(axis=1)]
four = four[~np.isnan(four).any(axis=1)]

data_country[:5]   = ['Afghanistan' 'Albania' 'Algeria' 'Angola' 'Argentina']
data_health[:5]    = [0.952 0.956 0.958 0.976 0.977]
data_education[:5] = [0.514 0.999 0.966 0.759 1.   ]
data_economy[:5]   = [0.18  0.748 0.456 0.646 0.639]
data_politics[:5]  = [0.132 0.377 0.151 0.245 0.39 ]

#4指標の散布図の関数
def show_scatter(country, four):
    jpidx = np.where(country=='Japan')[0][0]
    for i in range(4):
        plt.scatter(nr.randn(len(four))/50+i, four[:, i], color='k', s=2)
        plt.scatter([i], four[jpidx, i], color='r', marker='_', s=1000) #日本を赤線で
    plt.xticks(range(4), ['health', 'education', 'economy', 'politics'])
    plt.show()
    plt.close()

#4指標の散布図
show_scatter(country, four)

#ランキングの関数
def show_rank(country, val):
    #nanを除く
    c = country[~np.isnan(val)]
    v = val[~np.isnan(val)]
    #大きい順にソート
    idx = np.argsort(v)[::-1]
    c_sort = c[idx]
    v_sort = v[idx]
    
    jpidx = np.where(c_sort=='Japan')[0][0]
    print('japan rank = {}'.format(jpidx))
    
    #ベスト10とワースト10
    plt.subplot(121)
    plt.bar(range(10), v_sort[:10], color='gray', tick_label=c_sort[:10])
    plt.subplot(122)
    plt.bar(range(10), v_sort[-10:], color='gray', tick_label=c_sort[-10:])
    plt.show()
    plt.close()

#4指標の平均でジェンダーギャップ指数を計算
GGGI = np.mean(four, axis=1)
print('GGGI[:5] = {}'.format(GGGI[:5]))
#ランキングを表示
show_rank(country, GGGI)

GGGI[:5] = [0.4445  0.77    0.63275 0.6565  0.7515 ]
japan rank = 120

#4指標をそれぞれ偏差値にしてから散布図
four2 = deepcopy(four)
for i in range(4):
    four2[:, i] = 50 + 10 * (four2[:, i] - np.average(four2[:, i])) / np.std(four2[:, i])
show_scatter(country, four2)

#これでジェンダーギャップ指数を計算
GGGI_standard = np.mean(four2, axis=1)
print('GGGI_standard[:5] = {}'.format(GGGI_standard[:5]))
#ランキングを表示
show_rank(country, GGGI_standard)

GGGI_standard[:5] = [21.1705082  51.59715048 41.91055118 45.73514797 55.42146072]
japan rank = 109

#4指標をロジット関数に通してから散布図
four3 = deepcopy(four)
for i in range(4):
    #0除算回避
    four3[:, i] = np.clip(four3[:, i], 0.001, 0.999)
    #ロジット関数
    four3[:, i] = np.log(four3[:, i] / (1 - four3[:, i]))
show_scatter(country, four3)

#これでジェンダーギャップ指数を計算
GGGI_logit = np.mean(four3, axis=1)
print('GGGI_logit[:5] = {}'.format(GGGI_logit[:5]))
#ランキングを表示
show_rank(country, GGGI_logit)

GGGI_logit[:5] = [-0.08908965  2.6427489   1.14268642  1.08216416  2.69486537]
japan rank = 111

four4 = deepcopy(four3)
for i in range(4):
    #偏差値
    four4[:, i] = 50 + 10 * (four3[:, i] - np.average(four3[:, i])) / np.std(four3[:, i])
show_scatter(country, four4)

#これでジェンダーギャップ指数を計算
GGGI_logit_standard = np.mean(four4, axis=1)
print('GGGI_logit_standard[:5] = {}'.format(GGGI_logit_standard[:5]))
#ランキングを表示
show_rank(country, GGGI_logit_standard)

GGGI_logit_standard[:5] = [29.66993965 52.89487215 40.92859231 47.53691477 56.51405356]
japan rank = 114