ab_test_inference_solution.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "source": [
    "# 1. Постановка гипотезы\n",
    "\n",
    "Нужно определить, менять ли старый вариант работы сервиса standard на новый svm_ensemble_v_1_22.\n",
    "Посколько мы заранее не имеем представления о том, будет ли новый вариант работать лучше или хуже, чем текущий вариант, мы выбираем двусторонний статистический тест Two-Tailed Test\n",
    "H0 : p = p0\n",
    "H : p != p0\n",
    "где p и p0 означают коэффициенты конверсии между новым и старым вариантом работы сервиса соответственно. Так же необходимо выбрать уровень доверия (confidence level). Выбираем уровень доверия 95%. Имея уровень доверия в 95%, необходимо рассчитать показатель альфа, который будет равен (1 - 0.95) для нашего теста = 0.05."
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import scipy.stats as stats\n",
    "import statsmodels.stats.api as sms\n",
    "import matplotlib as mpl\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from math import ceil\n",
    "\n",
    "\n",
    "# Настройки отображения графиков\n",
    "%matplotlib inline\n",
    "\n",
    "plt.style.use('seaborn-whitegrid')\n",
    "font = {'family' : 'Helvetica',\n",
    "        'weight' : 'bold',\n",
    "        'size'   : 14}\n",
    "mpl.rc('font', **font)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Для каждой группы нам необходимо не меньше  4720  наблюдений\n"
     ]
    }
   ],
   "source": [
    "effect_size = sms.proportion_effectsize(0.13, 0.15)\n",
    "# Cчитаем необходимый размер выборки -> https://www.statsmodels.org/dev/generated/statsmodels.stats.power.NormalIndPower.html?highlight=normalindpower\n",
    "# https://www.statsmodels.org/dev/generated/statsmodels.stats.power.NormalIndPower.solve_power.html#statsmodels.stats.power.NormalIndPower.solve_power\n",
    "required_n = sms.NormalIndPower().solve_power(effect_size, power=0.8, alpha=0.05, ratio=1) # Calculating sample size needed\n",
    "# Округляем до ближайшего целого\n",
    "required_n = ceil(required_n)\n",
    "print(\"Для каждой группы нам необходимо не меньше \",required_n, \" наблюдений\")"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# 2. Сбор данных и их подготовка"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "outputs": [
    {
     "data": {
      "text/plain": "                                user_id                   timestamp  \\\n0  9109b0dc-d393-497f-8d63-ba9a25dd16b4  2022-05-21 22:11:48.556739   \n1  2430c3d2-f75b-4b31-8271-51b6a76c2652  2022-05-12 08:01:45.159739   \n2  44788c4e-8dd2-4fad-b986-75e76f4adb64  2022-05-11 16:55:06.154213   \n3  4699a417-506d-41b8-a354-6af6ad576963  2022-05-08 18:28:03.143765   \n4  304b0d28-bcdf-401a-9dff-66230d3ba0bc  2022-05-21 01:52:26.210827   \n5  426f77eb-8739-43b6-999c-aecd14d9f346  2022-05-10 15:20:49.083499   \n6  aa9d93c5-a18f-444a-a522-7e7145835362  2022-05-19 03:26:46.940749   \n7  de214e30-9b8d-44cf-87b1-16917f3e1c89  2022-05-17 01:48:29.539573   \n8  070f398e-e6bf-43e7-88bb-37eff2cb9212  2022-05-04 17:58:08.979471   \n9  9cff0b7e-d42b-4508-9fbe-a4e74f7f29c5  2022-05-15 18:11:06.610965   \n\n       group             variants  converted        location age_group  \\\n0    control             standard          0  United Kingdom     18-25   \n1    control             standard          0  United Kingdom     42-49   \n2  treatment  svm_ensemble_v_1_22          0  United Kingdom     26-33   \n3  treatment  svm_ensemble_v_1_22          0  United Kingdom     42-49   \n4    control             standard          1  United Kingdom     42-49   \n5    control             standard          0  United Kingdom     42-49   \n6  treatment  svm_ensemble_v_1_22          1  United Kingdom     26-33   \n7    control             standard          0  United Kingdom     18-25   \n8  treatment  svm_ensemble_v_1_22          1  United Kingdom     18-25   \n9  treatment  svm_ensemble_v_1_22          1  United Kingdom     26-33   \n\n         user_type  \n0  registered_user  \n1  registered_user  \n2         new_user  \n3  registered_user  \n4  registered_user  \n5  registered_user  \n6  registered_user  \n7  registered_user  \n8         new_user  \n9         new_user  ",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>user_id</th>\n      <th>timestamp</th>\n      <th>group</th>\n      <th>variants</th>\n      <th>converted</th>\n      <th>location</th>\n      <th>age_group</th>\n      <th>user_type</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>9109b0dc-d393-497f-8d63-ba9a25dd16b4</td>\n      <td>2022-05-21 22:11:48.556739</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>18-25</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>2430c3d2-f75b-4b31-8271-51b6a76c2652</td>\n      <td>2022-05-12 08:01:45.159739</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>42-49</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>44788c4e-8dd2-4fad-b986-75e76f4adb64</td>\n      <td>2022-05-11 16:55:06.154213</td>\n      <td>treatment</td>\n      <td>svm_ensemble_v_1_22</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>26-33</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>4699a417-506d-41b8-a354-6af6ad576963</td>\n      <td>2022-05-08 18:28:03.143765</td>\n      <td>treatment</td>\n      <td>svm_ensemble_v_1_22</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>42-49</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>304b0d28-bcdf-401a-9dff-66230d3ba0bc</td>\n      <td>2022-05-21 01:52:26.210827</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>1</td>\n      <td>United Kingdom</td>\n      <td>42-49</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>426f77eb-8739-43b6-999c-aecd14d9f346</td>\n      <td>2022-05-10 15:20:49.083499</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>42-49</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>6</th>\n      <td>aa9d93c5-a18f-444a-a522-7e7145835362</td>\n      <td>2022-05-19 03:26:46.940749</td>\n      <td>treatment</td>\n      <td>svm_ensemble_v_1_22</td>\n      <td>1</td>\n      <td>United Kingdom</td>\n      <td>26-33</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>de214e30-9b8d-44cf-87b1-16917f3e1c89</td>\n      <td>2022-05-17 01:48:29.539573</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>18-25</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>070f398e-e6bf-43e7-88bb-37eff2cb9212</td>\n      <td>2022-05-04 17:58:08.979471</td>\n      <td>treatment</td>\n      <td>svm_ensemble_v_1_22</td>\n      <td>1</td>\n      <td>United Kingdom</td>\n      <td>18-25</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>9cff0b7e-d42b-4508-9fbe-a4e74f7f29c5</td>\n      <td>2022-05-15 18:11:06.610965</td>\n      <td>treatment</td>\n      <td>svm_ensemble_v_1_22</td>\n      <td>1</td>\n      <td>United Kingdom</td>\n      <td>26-33</td>\n      <td>new_user</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv('./ab_test_inference/ab_test_ensembles.csv')\n",
    "df.head(10)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "outputs": [
    {
     "data": {
      "text/plain": "(294478, 8)"
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "outputs": [
    {
     "data": {
      "text/plain": "treatment    147276\ncontrol      147202\nName: group, dtype: int64"
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['group'].value_counts()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "outputs": [
    {
     "data": {
      "text/plain": "variants   standard  svm_ensemble_v_1_22\ngroup                                   \ncontrol      145274                 1928\ntreatment      1965               145311",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th>variants</th>\n      <th>standard</th>\n      <th>svm_ensemble_v_1_22</th>\n    </tr>\n    <tr>\n      <th>group</th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>control</th>\n      <td>145274</td>\n      <td>1928</td>\n    </tr>\n    <tr>\n      <th>treatment</th>\n      <td>1965</td>\n      <td>145311</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.crosstab(df['group'], df['variants'])"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "У нас имеются 0 пользователей, которые не уникальны в нашей выборке.\n"
     ]
    }
   ],
   "source": [
    "#все пользователи уникальны\n",
    "sessions = df['user_id'].value_counts(ascending=False)\n",
    "duplicate_users = sessions[sessions > 1].count()\n",
    "print(f\"У нас имеются {duplicate_users} пользователей, которые не уникальны в нашей выборке.\")"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "outputs": [
    {
     "data": {
      "text/plain": "0"
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "duplicate_users"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "outputs": [
    {
     "data": {
      "text/plain": "                                user_id                   timestamp    group  \\\n0  bcf8452b-3056-4d45-994f-5f286fd9869a  2022-05-18 04:44:18.590293  control   \n1  3c827a98-0868-4ce5-ac31-8195e243af37  2022-05-14 05:22:43.771772  control   \n2  71c72bb9-13d9-4755-b398-b2c5d8b1c4d6  2022-05-23 08:52:53.050325  control   \n3  654ee878-b0c5-4223-b9fb-995ca3d2eaf1  2022-05-07 23:27:09.109877  control   \n4  1e044366-7e86-4d92-bba7-3afaa6f887f8  2022-05-20 14:26:10.133979  control   \n\n   variants  converted        location age_group        user_type  \n0  standard          0  United Kingdom     42-49         new_user  \n1  standard          0  United Kingdom     50-57         new_user  \n2  standard          0  United Kingdom     34-41         new_user  \n3  standard          0  United Kingdom     26-33  registered_user  \n4  standard          0  United Kingdom     42-49         new_user  ",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>user_id</th>\n      <th>timestamp</th>\n      <th>group</th>\n      <th>variants</th>\n      <th>converted</th>\n      <th>location</th>\n      <th>age_group</th>\n      <th>user_type</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>bcf8452b-3056-4d45-994f-5f286fd9869a</td>\n      <td>2022-05-18 04:44:18.590293</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>42-49</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>3c827a98-0868-4ce5-ac31-8195e243af37</td>\n      <td>2022-05-14 05:22:43.771772</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>50-57</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>71c72bb9-13d9-4755-b398-b2c5d8b1c4d6</td>\n      <td>2022-05-23 08:52:53.050325</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>34-41</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>654ee878-b0c5-4223-b9fb-995ca3d2eaf1</td>\n      <td>2022-05-07 23:27:09.109877</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>26-33</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1e044366-7e86-4d92-bba7-3afaa6f887f8</td>\n      <td>2022-05-20 14:26:10.133979</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>42-49</td>\n      <td>new_user</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "control_sample = df[df['group'] == 'control'].sample(n = required_n, random_state=42)\n",
    "treatment_sample = df[df['group'] == 'treatment'].sample(n = required_n, random_state=42)\n",
    "\n",
    "# Сформируем общую случайную выборку\n",
    "ab_df = pd.concat([control_sample, treatment_sample], axis=0)\n",
    "ab_df.reset_index(drop=True, inplace=True)\n",
    "ab_df.head()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "outputs": [
    {
     "data": {
      "text/plain": "control      0.5\ntreatment    0.5\nName: group, dtype: float64"
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ab_df.group.value_counts(normalize=True)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# 3. Визуализация результатов эксперимента"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "outputs": [
    {
     "data": {
      "text/plain": "            CR mean    CR std     CR se\ngroup                                  \ncontrol    0.132415  0.338942  0.004933\ntreatment  0.119915  0.324862  0.004729",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>CR mean</th>\n      <th>CR std</th>\n      <th>CR se</th>\n    </tr>\n    <tr>\n      <th>group</th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>control</th>\n      <td>0.132415</td>\n      <td>0.338942</td>\n      <td>0.004933</td>\n    </tr>\n    <tr>\n      <th>treatment</th>\n      <td>0.119915</td>\n      <td>0.324862</td>\n      <td>0.004729</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conversion_rates = ab_df.groupby('group')['converted']\n",
    "\n",
    "# Стандартное отклонение для сформированной выборки\n",
    "std_proportion = lambda x: np.std(x, ddof=0)\n",
    "# Стандартная ошибка для сформированной выборки\n",
    "se_proportion = lambda x: stats.sem(x, ddof=0)\n",
    "\n",
    "conversion_rates = conversion_rates.agg([np.mean, std_proportion, se_proportion]).rename(columns={'mean':'CR mean', '<lambda_0>':'CR std', '<lambda_1>':'CR se'})\n",
    "conversion_rates"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "По выборочным статистикам два эксперимента +/- показывают себя одинаково, за исключением выборочного среднего коэффициента конверсии ~13.2% против ~11.9%\n",
    "Выходит, что с вариантом svm_ensemble_v_1_22 выборочный средний коэффициент конверсии даже ухудшается."
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 576x432 with 1 Axes>",
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg8AAAGlCAYAAAB9ShW3AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8qNh9FAAAACXBIWXMAAAsTAAALEwEAmpwYAABVnklEQVR4nO3deXxM1//H8VcsSSSWSBCNfZtQibW1tXYhQm2tfWlRpbV9JVVaqq2qLtZa2ipNq6jta18qUUoVtcSXqq1EYkkIEvuWSO7vD4/Mr9MsZsgweD8fDw937r3n3M+ZzE0+c8659zoZhmEgIiIiYqVsjzoAERERebwoeRARERGbKHkQERERmyh5EBEREZsoeRARERGbKHkQERERmyh5kKfW6dOn8fX1ZenSpY86FIfj6+vLqFGjHnUYDm3Hjh34+vqyY8eORx2K3Q0fPpxGjRo96jDs5klvnz3keNQByMOXkJBAaGgov/76KzExMRiGQfHixalfvz7du3fH29v7UYcoVoiLi2PRokU0adKEChUqPOpwxEFt3ryZP//8k4EDBz7qUOQJouThKXPgwAH69OnD1atXadmyJV27diVbtmwcOXKExYsXs379esLCwh51mA9FkSJF+PPPP8mR4/E8Dc6dO8e0adMoUqSIkodH4Pnnn+fPP/8kZ86cjzqUTG3evJl58+YpeZAs9Xj+1pT7cvXqVfr37w/A0qVLKVeunMX24OBgZs6c+ShCyzKGYXD79m1cXV3vua+TkxMuLi4PISrr3LhxAzc3t0cdhtxD6s8pW7ZsDvX5EXmYNOfhKbJgwQLOnDnDsGHD0iQOAHny5CE4ONhiXVhYGO3ataNSpUrUrFmT4OBgYmNjLfYZPnw4/v7+xMXF8dZbb1G1alVq1arF559/TnJyMgBJSUnUqFGDoUOHpjnurVu3qFatGsOHDzevS0xMZNq0aTRt2hQ/Pz/q1q3L2LFjuXnzpkXZ1LH5tWvX8tJLL+Hv78/atWsB2L59O127duX555+ncuXKNGnShNGjR5vLZjTn4fDhw/Tp04dq1apRpUoVunfvzu7duy32Wbp0Kb6+vuzcuZNPP/2UWrVqUaVKFfr3709CQkKGP4NUU6dOxdfXl7///puhQ4dSo0YNWrZsCUBMTAwfffQRzZs3p3Llyjz33HP07duXI0eOmMvv2LGDV155BYB3330XX19ffH19mTp1qnmfqKgoBg8eTM2aNfH396dNmzasW7funrH909q1awkKCsLf359WrVrx22+/mbdFR0fj6+vL999/n6bc4cOH8fX15aeffsq0fsMwmDNnjvlnV7t2bd577z2L9zD1vdqyZYtF2c8//5wKFSoQERFhfk98fX1ZuXIlU6ZM4cUXX6Ry5cr07t2b6OjoNMe25v1J/Tlv376dMWPGUKdOHapWrWpxvH/OeUg9F2JjY+nbty9Vq1blxRdf5McffwTg6NGjvPrqq1SpUoUGDRqwfPnyNHFdvXqVTz/9lAYNGuDn50fjxo2ZPn26+VyC///sfvvtt+ahKz8/P15++WX+/PNPi3jmzZsHYP6M+Pr6cvr0aXP7XnvtNV544QX8/Pxo2rQpM2bMICUlJdOfW0YuXLjAiBEjqF+/Pn5+ftSpU4fXX3+do0ePmvfZsGED/fr1o169evj5+dGwYUM+//xzbt++bVHXg76XqT+7P/74g9GjR1OzZk2qVq3K4MGDuXDhglXtWbVqFS+//DKVKlXi+eefZ9CgQZw6dcpinxMnTjB48GBefPFF/Pz8ePHFFxk4cCDnzp27j3fw8aGeh6fIxo0bcXFxoXnz5lbtv2LFCt555x0qVqxIcHAwFy9e5McffyQiIoJly5bh6elp3tcwDF5//XX8/f1555132L59O6GhoRQrVowuXbqQM2dOAgIC+Pnnn7l9+7bFN7bNmzdz/fp1goKCzHX179+fXbt20b59e8qWLUtkZCQ//fQTx44d47vvvsPJyclcfvfu3YSFhdGtWzcKFChA6dKlOXbsGG+88QYmk4kBAwaQK1cuTp48ye+//55pmyMjI+nSpQtubm707t0bFxcXFi9ezGuvvcb333/P888/b7H/2LFjyZcvHwMGDCAmJobZs2czevRoJk+ebNV7PGTIEIoWLcrgwYNJSkoCYP/+/ezevZtmzZrh4+PDuXPnWLBgAd27d2f16tUUKlSIMmXKMGjQIKZMmULHjh2pXr06cPcPRGo7OnXqRIECBejduzfu7u6sX7+ewYMH88UXX9C6det7xva///2PdevW0b17d9zd3Vm4cCFvvvkms2fP5rnnnqNkyZJUrVqVlStX0rNnT4uyK1euJGfOnOafaUY++OADlixZQps2bejatStnzpxh7ty57N+/n//+97+4uLjw5ptv8uuvvzJixAhWr15N3rx52b17Nz/88AO9evUytz3VzJkzSUlJoXfv3ly+fJkff/yRHj16sHLlSjw8PO7r/RkzZgx58+alX79+XL16NdM2GYbBG2+8QbVq1Rg6dCirVq3ik08+wd3dnS+//JIWLVoQEBDAggULePfdd6latSolSpQA7ibSPXr0ICYmhk6dOpmH1qZNm0ZsbCyffPKJxbHWrl3LjRs36NixI05OTsyaNYuBAwfyyy+/kDNnTjp27Mi5c+fYunUrX3zxhblc6rk7b948ypQpQ/369XF2duaPP/5g4sSJXL16lbfffjvTdqZn0KBB/P3333Tt2pWiRYty8eJFdu7cSXR0tPkLy9KlS3F2dqZ79+7kyZOHffv2MXv2bM6ePcukSZOy7L1M9cknn+Dm5kb//v2JjY1l7ty5REVF8d///hdnZ+cM2/Ltt98yceJEmjVrRrt27bhy5Qrz5s2jc+fOrFy5Ek9PT5KSkujduze3bt2iS5cuFCxYkPPnz7NlyxbOnTtHoUKFbH4PHxuGPDWef/55o1WrVlbtm5iYaNSpU8do3ry5cfPmTfP6P/74wzCZTMZnn31mXjds2DDDZDIZU6dOtaijTZs2Rtu2bc2vt27daphMJiM8PNxiv0GDBhk1a9Y0kpKSDMMwjJUrVxq+vr7Gjh07LPZbsWKFYTKZjC1btpjXmUwmw9fX1zh48KDFvj/88INhMpmM+Pj4DNt46tQpw2QyGUuWLDGv69+/v1GxYkUjKirKvC4+Pt6oUaOGRVuWLFlimEwm49VXXzVSUlLM68eOHWtUqFDBuHLlSobHNQzDmDJlimEymYwBAwak2fbP9zvVyZMnDT8/P2P69OnmdX/++Wea+FP17NnTCAoKSlNXz549jbp161rEnB6TyWSYTCZjz5495nUJCQnG888/b3Tq1Mm87qeffjJMJpNx9OhR87rk5GSjXr16Rv/+/TM9RkREhGEymYxly5ZZrN+1a5dhMpmMBQsWmNcdPXrU8Pf3N95++23j2rVrRuPGjY2WLVsat2/fNu+T+tmsXbu2cfnyZfP6bdu2GSaTyZg4caLN70/qz/mVV14xfz7/fbw//vjDvC71XJg2bZp53eXLl41KlSoZvr6+xooVK8zrIyMj08T19ddfG5UqVTKOHTtmcayvvvrKMJlMRmRkpGEY///ZrVGjhnHp0iXzfr/88othMpmMjRs3mtd99NFHhslkMtJz48aNNOtGjhxpVKlSxeK9HTZsmNGwYcN06/hnO00mkzFr1qxM90vvmF9//bXh6+trxMbGWhzzQd7L1J/dvz8nixcvNkwmk7Fo0aIM2xcTE2M8++yzaX6nnThxwvDz8zMmTJhgGIZhHDp0yDCZTMbPP/+caZufRBq2eIpcu3YNd3d3q/b966+/uHDhAp07d7aYP1CzZk0qVqzIpk2b0pTp0KGDxevq1aubu0dTyxYoUMA8rAB3x483b95M06ZNzRMXf/75Z0qWLEnZsmVJSEgw/6tRowZOTk5pLo2rWrVqmgmDefLkAe52kVrbBZucnMzvv/9Ow4YNKVmypHm9p6cn7dq148CBA2m6O1955RWLXpDnnnuO5ORkYmJirDpm586d06z75/t98+ZNLl68iLu7O6VKleLAgQP3rPPSpUts27aN5s2bc+PGDYv3sG7dusTFxREVFXXPevz8/Mxd9AD58+enZcuW7Nmzh8uXLwMQFBSEs7MzK1asMO+3Y8cOzp49e8/ejZ9//hk3Nzfq1q1rEWPp0qUpUKCAxc+5bNmyDBkyxNzLcfbsWT777LN0vzm2bt2avHnzml/Xrl2bcuXKmT+z9/P+dOjQwaaJte3btzcv582bl1KlSuHi4mIemgIoXbo0efPmtThHfv75Z6pXr07+/Pkt4qpTpw4AO3futDhOs2bNyJcvn/n1c889B5Cmaz0juXLlAu5+9i9fvkxCQgLPP/88N27c4Pjx41a3F+5+bnPmzMnOnTu5dOnSPY+ZkpLC1atXSUhIoHr16hiGke7n+37fy1SdOnWy+Jy0adOGvHnz8uuvv2YYY3h4OHfu3CEoKMji55A7d25MJpP5s5n6+/T333/nxo0bGdb3JNKwxVMkd+7cXL9+3ap9U+c1lCpVKs22MmXKpLkiI2fOnGm66PLly2f+IwOQPXt2mjVrxtKlS82TzjZu3MjNmzdp0aKFeb/o6GiioqKoXbt2urHFx8dbvC5evHiafYKCgvjvf//LyJEjGT9+PLVq1aJJkyY0b948wz8CCQkJ3Lx5M902ly5dGrg7H6FAgQLm9c8884zFfql/tK5cuZLuMf6tWLFiadbdvn2bL7/8kpUrV3L+/HmLband7pk5efIkhmEwdepUizkQ/5T6Rzoz/0yg/r0uNjaWfPnykS9fPho1asTq1asJDg7GycnJPDxQv379TOuPjo7mxo0b5j+M//bvn/Orr77KmjVr2LdvH/3796dixYo2xf3HH38A9/f+pPdzykh650KePHnw9vYmW7Zsadb/8xyJjo7m8OHDVn/2//35S00krP387d69m0mTJrFv3z7zsFmqew3P/JuzszNvv/02X3zxBS+88AKVKlWifv36tG7d2iLOv//+m3HjxrFz505u3bplUce1a9csXj/Ie5nq38MYOXLkoGjRopkm+KlzZDIa4k39PBQrVoyePXvy/fffs3LlSqpVq0bDhg1p1aoV+fPnz7D+J4GSh6dI6dKlOXjwIImJiZmO9d2Pf377zkxQUBDz5s1j06ZNBAUFsXbtWgoVKmQxlyAlJYWyZcsyYsSIdOv49y+T9Ga8u7q6MnfuXHbt2sXmzZv5/fffefvtt/n+++/56aefrLoawxrZs2dPd71hGFaVTy+Ojz/+mCVLltCtWzeqVq1K3rx5yZYtG2PHjrWq3tSeltdeey3DP+DpTZi9X6kTDXft2kXlypUJDw+nZcuW9/yMpaSk4OHhkWacO9U/ew/gbuIWGRkJ3P0DdL/u5/2x5fOS0bnw7z92qf75M01JSaFWrVr07ds33X3/ncQ8yOfv1KlT9OzZk5IlS/Luu+/i4+ODi4sLBw4cYPz48fc1afK1116jcePGbNiwgW3btvHVV18xY8YMvvnmG2rWrMnVq1fp0aMHbm5uDBkyhOLFi+Pq6kpcXBzDhw9Pc8wHeS8fRGocM2fOTPfLxj9/5wwfPpyXX36ZjRs38vvvv/P555/z9ddfM3fuXMqWLZsl8TgiJQ9PkUaNGpknwbVq1SrTfX18fIC7M9JffPFFi23Hjx+nSJEi9xVD9erVeeaZZ1i7di316tVjy5YtdOrUyeKXQfHixTlw4AC1a9e2OilJT7Zs2ahZsyY1a9bknXfe4aeffuKjjz4iPDw83fZ7enqSK1eudLv0U7tw77fdtli3bh1t2rRJkzxdvnzZ4ttMRu9N6h+Y7NmzZ/it3hrpXaGQui718wFQt25dvLy8WLFiBRcuXODatWtWTcgsXrw427Zto3LlyvccTjMMg/feew8XFxd69erFtGnTWLFiRbrHySju1J9dVr0/9lC8eHGuX7+epXFl9DnZsGEDiYmJfPPNNxaf6/S6/m1RrFgxXnvtNV577TXz8FVq8rBjxw4uXrzIlClTqFGjhrnM1q1bH+iYmTlx4oTF77A7d+5w+vTpNJOf/ym1N9PHx8eqBKBcuXKUK1eOvn37cvjwYV5++WV++OEHxowZ8+ANcFCa8/AU6dSpE97e3nz++efmb3D/dO3aNfO3QD8/PwoUKMDChQstLqHavXs3f/31Fw0aNLivGJycnGjevDm//fYby5cvJzEx0WLIAu72Tly4cCHdy/wSExPTdG2m5+LFi2nWpXZzZ9Qdmz17durWrcuvv/7KyZMnzesvXbrE8uXLze+JvWXPnj3NN6jVq1enufQrdez4313UXl5e1KxZk8WLFxMXF5emfmsuJYW7817+97//mV9fvHiR1atXU7VqVYtx9hw5cvDSSy8RFhbGkiVLKFGiBNWqVbtn/UFBQaSkpDB9+vQ021LH4FPNnj2bnTt38sEHHzBgwADq1KnDJ598km77VqxYYfGebN++naNHj5o/s1n1/thD8+bN2b9/P5s3b06z7dq1ayQmJtpcZ+rn5N9d+qm9Fv/8rCUmJjJ37lybjwF35+f8exiicOHCeHl5mX8eqV8S/t3bkt7lvlllwYIFFu/b8uXLuXLlSqa/w5o1a0b27NmZPn16ur0ZqZ+Ra9eucefOHYttZcqUwcXFxeqho8eVeh6eInnz5mX69Om88cYbtG3blpYtW+Lv70+2bNk4evQoq1evJl++fAwZMoScOXMydOhQhg0bRpcuXWjVqhUJCQnMmTMHb29v+vTpc99xBAUFERoaysSJEylSpAhVqlSx2N6qVSvWrVvH6NGj2bVrl3kyVVRUFD///DNffvklNWvWzPQYX331FTt37qRBgwYUKVKEy5cvs2DBAtzc3DL9pfGf//yH33//nS5dutClSxdcXFxYtGgRV69etbgPhT01atSI5cuXkzt3bsqVK8ehQ4f4+eef03RZFy9enHz58jF//nzc3Nxwd3enXLlymEwmPvzwQzp37kyrVq1o3749xYsXJz4+nn379hEZGcn69evvGYfJZKJfv35069bNfKnm9evXCQkJSbNvmzZt+OGHH/j999+tvpPh888/T9euXfnuu+84cuQIdevWJWfOnJw8eZKwsDAGDRpEu3btOH78OJMmTaJFixbmMeixY8fy0ksvMXLkyDQ3NvPy8qJz58688sorXLlyhdmzZ1OwYEFee+018z5Z8f7Yw+uvv86vv/7KW2+9RZs2bahYsSK3b9/m77//Zt26daxatYqiRYvaVKefnx8Ao0ePpl69euTIkYOGDRvy4osvkjNnTvr160fHjh1JTExkxYoVGQ4J3Et0dDSvvvoqgYGBlC1bFmdnZzZv3kxkZCTDhg0DoFq1anh4eDB8+HC6detGjhw5CAsLs/tkwx49etCiRQtiYmKYO3cuJpOJNm3aZLh/sWLFCAkJ4YsvviA2NpbGjRubJ2Ru2LCBoKAgBg4cyB9//MFHH31Es2bNzHOl1q5da3Hp+ZNKycNTxt/fn9WrV5ufbbFmzRoMw6BEiRJ07NiR7t27m/dt06YNuXLlYsaMGYwfP55cuXJRv3593n77bYt7PNxPDCVKlODEiRN06dIlzfZs2bIxbdo0Zs+ezfLly9mwYQOurq7me0ak3ssgM40bN+bMmTMsW7aMhIQEPDw8qFq1Kv3798906KFMmTLMnz+fCRMmMHPmTAzDwM/PjzFjxmTazZmVRowYQY4cOczX8Pv5+TFz5kzGjRtnsV/OnDn54osvmDBhAqNHjyYpKYkBAwZgMpkoXbo0S5YsYfr06SxfvpxLly6RP39+ypcvz+DBg62KI/VmX1OnTuXUqVOUKlWK6dOnp/s+VKhQAV9fX44cOXLPIbF/GjVqFM8++ywLFixg0qRJZM+eHR8fH5o3b06tWrVITk5m2LBh5MmTx+JBXc888wwjRoxg+PDhLFq0yOJKnz59+hAVFcV3333HlStXeO655xg5cqTFkE9WvD/24Orqypw5c5gxYwbr1q1jxYoVuLu7U7JkSd566y0KFixoc51NmzY1TzZNPd83bNhg/nlOmjSJcePGkT9/ftq0aUONGjXo1auXzccpXLgwL730Etu3b2fVqlU4OTlRsmRJPvnkE/MNzTw8PPj222/57LPPmDp1Km5ubjRt2tScyNnDiBEjCAsLY9q0aSQmJtKoUSNGjhx5zzk5vXv3pkSJEvzwww98/fXXGIaBt7c3tWrVIjAwELh7X5V69erx22+/sXjxYlxcXChbtizTp0+nSZMmdmmPo3AysmqGiYg81V5++WVy5szJggULHsnxd+zYQY8ePZg4cWKaoTB5+ixdupR3332XhQsXpundlAenOQ8i8sAOHTrEX3/9Rdu2bR91KCLyEGjYQkTu299//82BAwf44Ycf8PLyslvXs4g4FvU8iMh9CwsL49133+XWrVtMnDjRPLNfRJ5smvMgIiIiNlHPg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2CTHow7gcREREfGoQxAREXmoqlevnu56JQ82yOhNFBERedJk9qVZwxYiIiJiEyUPIiIiYhMlDyIiImITJQ8iIiJiEyUPIiIiYhMlDyIiImITJQ8iIiJiEyUPIiIiYhMlDyIiImITJQ8iIiJiEyUPIiIiYhMlDyIiImITJQ8iIiJiE4dOHtasWUPbtm2pVKkSNWrUYNCgQZw4cSLTMuHh4fTs2ZMaNWrg6+uLr68vS5cuTXffiIgI3njjDWrUqIGfnx/169dn6NCh9miKiIjIE8Nhk4fFixcTHBzMwYMHKViwIMnJyYSFhdG5c2fOnz+fYbldu3axa9cu8ufPn2n9YWFhdO/enc2bN5OSkkLZsmVxdnZm48aNWd0UERGRJ4pDJg+JiYlMmDABgGbNmrFhwwbWrl2Lu7s78fHxzJgxI8Oyffv2Zc+ePUydOjXDfW7evMmoUaNITk6mTZs2bN26leXLl7N+/Xo2b96c5e0RERF5kjhk8rB//34uXrwIQNOmTQHw9vamSpUqAGzZsiXDsgUKFMDZ2TnT+rdu3cqlS5cAuHPnDk2bNuW5556jZ8+enD59+sEbICIi8gTL8agDSM/Zs2fNy15eXmmWY2NjH6j+qKgo8/Lq1aspXbo0V65cYdu2bXTv3p3Vq1fj7e2dptyhQ4ce6LgiIiJPAodMHjJiGEaW1JOcnGxe7tChAx9//DGRkZEEBQVx5coVVqxYwRtvvJGmXIUKFbLk+CIiIo4uIiIiw20OOWxRuHBh83J8fLx5OSEhAQAfH58Hqv+fvQp+fn4AlClTBjc3NwANXYiIiGTCIZMHf39/PDw8gLuXXgLExcWxd+9eAOrWrQtAYGAggYGBzJ0716b6a9euTfbs2QE4cOAAAMePH+fGjRsAlCpV6kGbICIi8sRyyOTB2dmZ4OBg4O4llY0bNyYoKIjr16+TP39++vTpA9yduxAVFWWeXAnw448/EhAQYDHsMH78eAICAhg3bhxwt2ejR48eACxcuJCgoCBeeeUV4G6vxssvv/xQ2ikiIvI4ctg5Dx07diRXrlyEhoYSGRmJi4sLAQEBhISEpDuZMdXly5c5efKkxbr4+Hjzv1TDhg2jUKFCLFq0iJMnT1KoUCGCgoIYMmQIefPmtVu7REREHndORlbNQnzCRUREUL169UcdhoiIyEOR2d89hxy2EBEREcel5EFERERsouRBREREbKLkQURERGyi5EFERERsouRBREREbKLkQURERGyi5EFERERsouRBREREbKLkQURERGyi5EFERERsouRBREREbKLkQURERGyi5EFERERsouRBREREbKLkQURERGyi5EFERERsouRBREREbKLkQURERGyi5EFERERsouRBREREbKLkQURERGyi5EFERERsouRBREREbKLkQURERGyi5EFERERsouRBREREbOLQycOaNWto27YtlSpVokaNGgwaNIgTJ05kWiY8PJyePXtSo0YNfH198fX1ZenSpRnuf/bsWYt9f/3116xuhoiIyBPFYZOHxYsXExwczMGDBylYsCDJycmEhYXRuXNnzp8/n2G5Xbt2sWvXLvLnz3/PY6SkpPDOO+9w+fLlrAxdRETkieaQyUNiYiITJkwAoFmzZmzYsIG1a9fi7u5OfHw8M2bMyLBs37592bNnD1OnTr3ncWbNmsWOHTto3rx5lsUuIiLypHPI5GH//v1cvHgRgKZNmwLg7e1NlSpVANiyZUuGZQsUKICzs/M9j3HgwAGmTJlCw4YN6dy584MHLSIi8pRwyOTh7Nmz5mUvL680y7GxsQ9U/82bNwkJCcHDw4OxY8c+UF0iIiJPmxyPOgBbGIaRJfVMmDCB6OhovvvuOzw9Pa0ud+jQoSw5voiIyOPMIZOHwoULm5fj4+PNywkJCQD4+Pg8UP1HjhwBYMCAAQAkJyebtw0ePJgmTZowceLENOUqVKjwQMcVERF5XERERGS4zSGHLfz9/fHw8ADuXnoJEBcXx969ewGoW7cuAIGBgQQGBjJ37lybj2EYBjdu3ODGjRvcvn3bvP727dvcunXrwRogIiLyBHPIngdnZ2eCg4MZNWoUYWFhNG7cmEuXLnH9+nXy589Pnz59AIiKigIwT64E+PHHH5kzZw5JSUnmdePHj+frr7+madOmDB06lDlz5lgcb8eOHfTo0QOAb775hoYNG9q7iSIiIo8th0weADp27EiuXLkIDQ0lMjISFxcXAgICCAkJwdvbO8Nyly9f5uTJkxbr4uPjzf9ERETkwTgZWTUL8QkXERFB9erVH3UYIiIiD0Vmf/cccs6DiIiIOC4lDyIiImITJQ8iIiJiEyUPIiIiYhMlDyIiImITJQ8iIiJiEyUPIiIiYhMlDyIiImITJQ+P2O2k5HvvJPIY0GdZ5OnhsLenflq45MxO9aE/PuowRB5YxLgejzoEEXlI1PMgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIja557Mtrly5Qnh4OLt27SImJoZbt27h6enJs88+S7169ahWrdrDiFNEREQcRIbJQ1xcHFOmTGHVqlUUKlSISpUq4efnh4uLC5cvX2bXrl388MMP+Pj4MGDAAIKCgh5m3CIiIvKIZJg8tGnThjZt2rBkyRLKlSuX7j63bt3il19+4fvvv+fMmTP07t3bboGKiIiIY8gweVizZg2enp6ZFnZ1daVly5a0bNmShISELA9OREREHE+GEybvlTg86P4iIiLyeLrnhMlUZ8+eZdeuXcTHx2MYhsW2nj17ZnlgIiIi4pisSh5WrlzJe++9R44cOdL0MDg5OSl5EBEReYpYlTxMmTKFXr16MXjwYLJnz27vmMzWrFnDrFmziIyMxNXVlVq1ahESEkKJEiUyLBMeHs78+fM5cOAAly9fBuDTTz+lXbt25n3i4uL46quv2LNnD3Fxcdy+fZtnnnmGFi1a0KtXL9zd3e3eNhERkceVVTeJio+Pp3379g81cVi8eDHBwcEcPHiQggULkpycTFhYGJ07d+b8+fMZltu1axe7du0if/78Ge4THR3NggULOHnyJD4+Pri6uhIVFcW0adMIDg62R3NERESeGFYlD/Xq1WPfvn32jsUsMTGRCRMmANCsWTM2bNjA2rVrcXd3Jz4+nhkzZmRYtm/fvuzZs4epU6dmuE++fPn4+OOP2blzJ8uXL2fz5s1UrlwZgE2bNpl7LERERCQtq4YtXnjhBcaPH8/Ro0fx9fUlRw7LYk2bNs3SoPbv38/Fixct6vb29qZKlSps3bqVLVu2ZFi2QIEC96y/fPnylC9f3vza1dWVatWqsW/fPrJly5amfSIiIvL/rPorOWrUKIB0v/E7OTlx6NChLA3q7Nmz5mUvL680y7GxsVl6vHPnzrF69WoAWrVqleGch6xuJ0CFChWyvE6RR8Ue54g9lS1VnJyumuMkj7ekW9c5FnXyoR7TquTh8OHD9o7DKv++RDQrHDt2jL59+3L+/Hmee+45Pvjggwz31R96kcw9jufIydH+jzoEkQdSfNR+u5x7ERERGW5zyKdqFi5c2LwcHx9vXk69i6WPj0+WHGfLli106tSJ06dPExAQwHfffYebm1uW1C0iIvKksjp52LRpE127dqVmzZrUqlWLbt26sXnzZrsE5e/vj4eHB3D30ku4e3nl3r17Aahbty4AgYGBBAYGMnfuXJuPMW/ePPr27cvVq1fp3bs3U6ZMwdXVNUviFxEReZJZlTwsXryYAQMGULx4cd5++21CQkIoWrQo/fv357///W+WB+Xs7Gy+ZDIsLIzGjRsTFBTE9evXyZ8/P3369AEgKiqKqKgo8+RKgB9//JGAgADeeOMN87rx48cTEBDAuHHjANi7dy+jR48mOTkZV1dXdu/eTadOnejQoQMdOnTg3LlzWd4mERGRJ4VVcx5mzpzJ8OHD6datm3ld+/btqVixIjNnzuSVV17J8sA6duxIrly5CA0NJTIyEhcXFwICAggJCcHb2zvDcpcvX+bkScuJI/Hx8eZ/ALdv3zZvu3XrVprLUBMTE7OwJSIiIk8Wq5KH2NhY81DBP9WrV4/PP/88y4NK1apVK1q1apXh9iNHjqRZN3DgQAYOHJhpvTVr1ky3rIiIiNybVcMWPj4+bN26Nc3633//nSJFimR5UCIiIuK4rOp56NWrF2PGjOHgwYNUrVoVgD179rBixQref/99uwYoIiIijsWq5KFTp054eXkRGhrK+vXrAShdujSTJ0+mSZMmdg1QREREHIvV92EOCAggICDAnrGIiIjIY8AhbxIlIiIijivDnodq1arxyy+/4OnpSdWqVXFycsqwkj179tglOBEREXE8GSYP77//Prlz5zYvZ5Y8iIiIyNMjw+Shbdu25uV27do9lGBERETE8Vk156Fx48YWt4BOdeXKFRo3bpzlQYmIiIjjsip5iImJISUlJc36xMRE4uLisjwoERERcVyZXqqZ+kRLuPtUzTx58phfJycns337dt1hUkRE5CmTafIwaNAg8/KIESMsC+bIQZEiRRg+fLh9IhMRERGHlGnycPjwYQAaNWrEkiVLyJ8//0MJSkRERBzXPec8JCUl4eXlle6ESREREXn63DN5yJkzJ6dPn9Z9HkRERASw8mqLNm3asGjRInvHIiIiIo8Bqx6MdfPmTVatWsW2bduoWLEibm5uFttHjhxpl+BERETE8ViVPERGRvLss88CcOrUKYttGs4QERF5uliVPMyZM8fecYiIiMhjwqrkIdXt27c5ceIETk5OFC9eHBcXF3vFJSIiIg7KquQhKSmJiRMnMm/ePJKSkjAMA2dnZ7p168aQIUPImTOnveMUERERB2FV8jB+/HjWrFnDRx99RPXq1QHYvXs3EydOxDAMhg0bZtcgRURExHFYlTysXr2asWPHUr9+ffO64sWL4+npyciRI5U8iIiIPEWsus/D1atXKVasWJr1xYoV48qVK1kelIiIiDguq5KH8uXLp3vFxY8//kiFChWyPCgRERFxXFYNWwwdOpQ33niDbdu2UaVKFQD27t3LuXPnmDlzpj3jExEREQdjVc/D888/z7p16wgMDOTGjRvcuHGDwMBA1q1bx3PPPWfvGEVERMSBWH2fB29vb4YMGWLPWNJYs2YNs2bNIjIyEldXV2rVqkVISAglSpTIsEx4eDjz58/nwIEDXL58GYBPP/2Udu3aWex37do1pkyZwrp160hISMDb25s2bdrQr18/XXoqIiKSCauTh3PnzjF//nwiIyMBKFOmDJ06dcLb29sugS1evNj8zIyiRYty6dIlwsLC2L17NytWrKBgwYLpltu1axe7du2iSJEi5uTh31JSUnjzzTfZuXMnOXPmpGjRopw4cYJp06Zx6tQpvvjiC7u0SURE5Elg1bDF1q1bCQgIYO3atbi6uuLq6srPP/9M06ZN+f3337M8qMTERCZMmABAs2bN2LBhA2vXrsXd3Z34+HhmzJiRYdm+ffuyZ88epk6dmuE+v/zyCzt37gRg6tSprFu3jvfeew+AFStWcODAgSxsjYiIyJPFquRhzJgxtG/fnnXr1vHFF1/wxRdfsG7dOtq3b88nn3yS5UHt37+fixcvAtC0aVPg7rBJ6mTNLVu2ZFi2QIECODs7Z1r/b7/9BoCrq6v53hWpx7lX/SIiIk87q4YtYmJi6Nq1a5onaHbt2pXFixdneVBnz541L3t5eaVZjo2NfaD6z5w5A4CHhwfZst3NnwoUKGDenlH9hw4deqDjpkeXusqTxB7niD3p/JMnxcM+96xKHvz8/Pj7778pVaqUxfq///77oZ58hmE80rr1i0YkczpHRB4Ne5x7ERERGW6zKnno0qULn376KSdOnKBy5coA7Nu3j/nz5xMSEmIxR6BixYoPGC4ULlzYvBwfH29eTkhIAMDHx+eB6n/mmWcAuHjxIikpKWTLls3iOA9av4iIyJPMquTh7bffBmDixIkZbgNwcnLKkq4Tf39/PDw8uHTpEuHh4bRs2ZK4uDj27t0LQN26dQEIDAwEoFu3bnTr1s3q+uvWrcvixYu5ffs2mzdvpmHDhoSHh1tsFxERkfRZlTxs2LDB3nFYcHZ2Jjg4mFGjRhEWFkbjxo25dOkS169fJ3/+/PTp0weAqKgoAPPkSrh7y+w5c+aQlJRkXjd+/Hi+/vprmjZtytChQ2nSpAnVq1cnIiKCgQMHUqxYMaKjowFo2bJllvSeiIiIPKmsSh6KFCli7zjS6NixI7ly5SI0NJTIyEhcXFwICAggJCQk03tLXL58mZMnT1qsi4+PN/8DyJ49O99++y2TJ08mLCyMU6dO8cwzz9C6dWveeustu7ZLRETkcedkWDkL8fDhw4SGhnLs2DGcnJwoW7YsvXv3xmQy2TtGhxAREUH16tXtUnf1oT/apV6RhyliXI9HHcJ9OTna/1GHIPJAio/ab5d6M/u7Z9V9HjZs2EC7du04c+YM9erVo27dusTGxtK2bVs2btyYpcGKiIiIY7Nq2GLy5Mn069ePQYMGWaz/8ssvmTx5Mo0aNbJLcCIiIuJ4rOp5iI6OpnXr1mnWt27d2jxpUURERJ4OViUPXl5e6T7v4cCBAxZ3ZhQREZEnn1XDFu3bt2fUqFGcOHGCqlWrArBnzx5CQ0Pp3bu3XQMUERERx2JV8vDWW2/h7u5OaGgoX375JQCFChVi4MCB9OjxeM6wFhERkftzz+Thzp07LFy4kObNm/Paa69x7do1AHLnzm334ERERMTx3HPOQ44cORg3bhx37twB7iYNShxERESeXlZNmKxcuXK6EyZFRETk6WPVnIcOHTrw+eefExsbi5+fH7ly5bLYrmdBiIiIPD2sSh5CQkIA+Oyzz9Jsy6onaYqIiMjjwSGfqikiIiKOy2GfqikiIiKOyarkAe7eTXL27NkcO3YMgDJlyvDaa69pvoOIiMhTxqqrLVauXMkrr7zC+fPnqV+/PvXr1yc+Pp727duzYsUKe8coIiIiDsTqp2oOHjyYfv36WayfMWMGX375ZboPzRIREZEnk1U9DwkJCTRv3jzN+sDAQOLj47M8KBEREXFcViUPNWvWZOfOnWnW79y5k+effz7LgxIRERHHZdWwRb169ZgwYQL79++nSpUqAOzdu5f169czYMAAwsPDzfs2bdrULoGKiIiIY7Aqefj4448BWLRoEYsWLUp3G+iGUSIiIk8Dq5KHw4cP2zsOEREReUxYNedBREREJFWGycOKFSswDMOqSmJiYti9e3eWBSUiIiKOK8PkYcmSJQQGBvLNN99w5MiRNInEpUuX2LBhA4MHD+aVV17h6tWrdg9WREREHr0M5zz8+OOPbNq0iTlz5vDll1/i4uKCp6cnLi4uXLlyhYSEBLy8vGjXrh0ffPABnp6eDzNuEREReUQynTDZoEEDGjRoQEJCAnv27CEmJobbt2+TP39+KlSowLPPPku2bJo2ISIi8jSx6moLT09PmjRpYu9YRERE5DFg9VM1H4U1a9Ywa9YsIiMjcXV1pVatWoSEhFCiRIlMy82ZM4f58+dz8uRJ8uTJQ4MGDQgJCaFAgQLmfbZv3863337LkSNHuHLlCh4eHlSuXJk333wTPz8/ezdNRETkseWwYw6LFy8mODiYgwcPUrBgQZKTkwkLC6Nz586cP38+w3KTJ09mzJgxREZGUqRIEW7cuMHSpUvp1q0bN2/eBCAyMpI+ffqwbds27ty5Q7ly5bh06RK//PILr776Krdv335YzRQREXnsOGTykJiYyIQJEwBo1qwZGzZsYO3atbi7uxMfH8+MGTPSLXfhwgVmzZoFQK9evQgLC2PhwoU4OTkRFRXFggULANi3bx9JSUnA3SeDLlu2jIEDBwJw/fp1Ll++bO8mioiIPLYcMnnYv38/Fy9eBP7/WRne3t7m52ps2bIl3XLbtm0zJwWp5cqXL28e5kgtV7lyZXLmzAlA3759adu2LVOnTsXd3Z2hQ4dSqFAh+zRMRETkCeCQycPZs2fNy15eXmmWY2NjH6hcmTJlCA0NxdPTk8uXL3Pw4EGSkpLw9vamXLlyWdcQERGRJ1CGEyanTZtmdSUDBgzIkmDuxdo7Xt6rXFxcHCNGjCAhIYEvvviCgIAAvvrqK2bOnMlbb73Fzz//TLFixdLUY4+HflWoUCHL6xR5VB63B+Pp/JMnxcM+9zJMHsLCwixex8TEcOvWLXOX/rlz53B1daVIkSJZnjwULlzYvBwfH29eTkhIAMDHx8eqcsWLF0+33E8//cTJkyfJlSsXrVu3BqBt27bMnDmTpKQkdu3alW7yoF80IpnTOSLyaNjj3IuIiMhwW4bDFqtWrTL/e+2116hYsSK//PILmzZtYtOmTfzyyy/4+/vz6quvZnnA/v7+eHh4ABAeHg7c7S3Yu3cvAHXr1gUgMDCQwMBA5s6dC0Dt2rXJkSOHRbnDhw9z4sQJi3Kpt9K+efMmR48eBe7Os0iVK1euLG+TiIjIk8KqOQ/Tp09nxIgRFt/4fXx8GD58ONOnT8/yoJydnQkODgbu9oA0btyYoKAgrl+/Tv78+enTpw8AUVFRREVFmSdXFixYkN69ewMQGhpKs2bN6NixI4ZhULJkSTp27AhAQEAATk5OALRv356XXnqJ9957z1xHapIhIiIiaVmVPMTHx3Pr1q0062/fvm3+w53VOnbsyLhx46hQoQLnzp3DycmJgIAA5s+fj7e3d4blhgwZwnvvvUfp0qWJiYkhV65ctGnThjlz5uDm5gbc7aH49ttvqVOnDu7u7kRHR1O4cGHatm3L/PnzyZ07t13aJCIi8iSw6g6TderU4f333+fjjz/G398fJycn/vzzT0aNGkWdOnXsFlyrVq1o1apVhtuPHDmSZp2TkxOvvvrqPYdT6tWrR7169R44RhERkaeNVcnDJ598wrBhw+jUqRPZs2cHICUlhRdffJExY8bYNUARERFxLFY/GGvmzJlERUVx/PhxAEqXLk2pUqXsGpyIiIg4HpsejFWqVCny5MmDp6enHsUtIiLylLIqA0hKSuKLL76gatWq1KtXj5iYGADGjRvHvHnz7BqgiIiIOBarkodp06bx66+/Mm7cOJydnc3rK1WqxLJly+wWnIiIiDgeq4Yt1qxZw9ixY6lRo4b5/ggA5cqVIzo62l6xiYiIiAOyqufh3Llz6d4SOjk5meTk5CwPSkRERByXVclD2bJl2b17d5r1P//8MxUrVszyoERERMRxWTVsMWDAAIYOHcqZM2dISUnh559/JioqilWrVvHtt9/aO0YRERFxIFb1PDRq1IjJkyezdetWsmXLxvTp04mOjuabb76x6x0mRURExPFYfZ+HunXr6oFRIiIiYl3PQ+PGjdN9ANaVK1do3LhxlgclIiIijsuq5CEmJoaUlJQ06xMTE4mLi8vyoERERMRxZTpsER4ebl7etGkTefLkMb9OTk5m+/btFClSxH7RiYiIiMPJNHkYNGiQeXnEiBGWBXPkoEiRIgwfPtw+kYmIiIhDyjR5OHz4MHD3aoslS5aQP3/+hxKUiIiIOK57znlISkrCy8sr3QmTIiIi8vS5Z/KQM2dOTp8+bfFMCxEREXl6WXW1RZs2bVi0aJG9YxEREZHHgFU3ibp58yarVq1i27ZtVKxYETc3N4vtI0eOtEtwIiIi4nisSh4iIyN59tlnATh16pTFNg1niIiIPF2sSh7mzJlj7zhERETkMWHVnIdUCQkJ7Nu3j8TERHvFIyIiIg7OquTh2rVrDBo0iDp16tCpUyfzLalHjRrF1KlT7RqgiIiIOBarkofx48dz7tw5li1bhqurq3l9w4YNWb9+vd2CExEREcdj1ZyHjRs3Mm3aNCpUqGCxvkyZMmkmUIqIiMiTzaqehytXrqR7a+rr16+TPXv2LA9KREREHJdVyYO/vz8bNmxIs37BggVUrVo1y4MSERERx2XVsMWQIUPo3bs3x44dIzk5mR9++IGjR4+yf/9+5s6da7fg1qxZw6xZs4iMjMTV1ZVatWoREhJCiRIlMi03Z84c5s+fz8mTJ8mTJw8NGjQgJCSEAgUKWOz366+/EhoayoEDB0hKSqJQoUI0a9aMd955x25tEhERedxZ1fNQrVo1FixYQFJSEsWLF2f79u0UKlSIBQsWULFiRbsEtnjxYoKDgzl48CAFCxYkOTmZsLAwOnfuzPnz5zMsN3nyZMaMGUNkZCRFihThxo0bLF26lG7dunHz5k3zfj/88AP9+vVj586d5MyZk7Jly2IYBlu2bLFLe0RERJ4UVvU8APj6+vL555/bMxazxMREJkyYAECzZs2YMmUKcXFxNG/enPj4eGbMmJHuLbEvXLjArFmzAOjVqxfDhg3j8OHDtGnThqioKBYsWEDPnj05e/Ys48ePB6Bv374MHjzYPHfj2rVrD6WNIiIijyureh5at25NaGgo586ds3c8AOzfv9/8CPCmTZsC4O3tTZUqVQAy7B3Ytm0bSUlJFuXKly9vHuZILRcWFkZSUhJOTk6cO3eOevXqUbNmTQYMGKBHj4uIiNyDVT0P9evXZ+7cuUyYMIHnn3+e1q1b07RpU9zd3e0S1NmzZ83LXl5eaZZjY2NtLhcdHW0uFxUVBYBhGCxfvpyyZcty8uRJ1q9fz8GDB1m1alW6bTt06NADtCp9/778VeRxZo9zxJ50/smT4mGfe1YlD8HBwQQHB7N7925Wr17N559/zkcffUTDhg1p3bo1DRo0sHOYdxmGkSXlkpOTzcv/+c9/6NevH7/99ht9+vQhJiaGTZs20aJFizT16BeNSOZ0jog8GvY49yIiIjLcZtOzLZ577jk+/PBDfv/9dyZNmkRUVBRvvvnmAwf4b4ULFzYvx8fHm5cTEhIA8PHxeaBy3t7e5m1+fn7A3ctRU8XExNx37CIiIk86m5IHgDNnzhAaGsrEiRM5cuQI1atXz/Kg/P398fDwACA8PByAuLg49u7dC0DdunUBCAwMJDAw0Hy5aO3atcmRI4dFucOHD3PixAmLci+88IL5WAcPHgTgr7/+Mq8rWbJklrdJRETkSWHVsMXly5dZt24dq1atYs+ePZQqVYpWrVrx0ksvZdgL8CCcnZ0JDg5m1KhRhIWF0bhxYy5dusT169fJnz8/ffr0Af5/7kLqJMeCBQvSu3dvZsyYQWhoKBs3buTs2bMYhkHJkiXp2LEjAFWrVqVp06aEh4czadIkVq1aZU4wKlasSKNGjbK8TSIiIk8Kq5KHF154AU9PT4KCgnjvvfd49tln7R0XHTt2JFeuXISGhhIZGYmLiwsBAQGEhIRYDDv825AhQ/Dy8mLBggWcOnWK3LlzExgYSEhICG5ubub9JkyYwPTp01m1ahVRUVH4+PjQuHFj+vfvb+69EBERkbScDCtmIW7dupXatWuTLZvNoxxPjIiICLsM0QBUH/qjXeoVeZgixvV41CHcl5Oj/e+9k4gDKz5qv13qzezvntU9DyIiIiJgZfJw6dIlJk2axB9//EF8fDwpKSkW2/fs2WOX4ERERMTxWJU8jBgxgkOHDtGhQwcKFSqEk5OTveMSERERB2VV8rB9+3a+//57KleubO94RERExMFZNQPSy8vL4koFEREReXpZlTwMGTKEKVOmcP36dXvHIyIiIg7OqmGLr7/+mtOnT1OnTh18fHzS3Adh1apVdglOREREHI9VyUOzZs3sHYeIiIg8JqxKHgYMGGDvOEREROQxYdN9mLdv305kZCROTk6ULVuWmjVr2isuERERcVBWJQ9xcXH079+fAwcOUKhQIQDOnTuHn58f06ZNy/RZEyIiIvJksepqizFjxpA9e3bCw8PZvHkzmzdvJjw8nOzZs/PJJ5/YO0YRERFxIFYlD1u3bmXUqFEUK1bMvK5YsWKMGDGCrVu32i04ERERcTxWPyYzvVtS6zbVIiIiTx+rkofatWvz8ccfc+bMGfO62NhYxo4dS+3ate0WnIiIiDgeqyZMjhw5kjfffJMmTZpYTJg0mUyMHDnSrgGKiIiIY7EqeXjmmWdYtmwZ27Zt4/jx4wCUKVOGOnXq2DU4ERERcTxW3+fBycmJF154gRdeeMGe8YiIiIiDy3TOw+bNm2nUqBHXrl1Ls+3q1as0atRIV1uIiIg8ZTJNHubNm0fv3r3JnTt3mm158uTh9ddfZ/bs2XYLTkRERBxPpsnDkSNHMr2aolatWhw+fDjLgxIRERHHlWnykJCQQLZsGe/i5OTEpUuXsjomERERcWCZJg+FCxfmyJEjGW4/cuSInmshIiLylMk0eahfvz5ffvklt27dSrPt5s2bTJkyhfr169stOBEREXE8mV6q+eabbxIWFkazZs3o2rUrpUuXBuD48ePMnTsXwzDo16/fQwlUREREHEOmyYOXlxcLFizgww8/ZNKkSRiGAdyd6/Diiy8yatQoChQo8FACFREREcdwz5tEFSlShJkzZ3L58mVOnDgBQIkSJciXL5/dgxMRERHHY/VTNfPly0elSpWoVKnSQ0sc1qxZQ9u2balUqRI1atRg0KBB5gQmM3PmzCEoKAg/Pz9q167Nu+++y4ULF9Ld9+DBg/j5+eHr64uvry9///13VjdDRETkiWJ18vCwLV68mODgYA4ePEjBggVJTk4mLCyMzp07c/78+QzLTZ48mTFjxhAZGUmRIkW4ceMGS5cupVu3bty8edNi31u3bhESEkJSUpK9myMiIvLEcMjkITExkQkTJgDQrFkzNmzYwNq1a3F3dyc+Pp4ZM2akW+7ChQvMmjULgF69ehEWFsbChQtxcnIiKiqKBQsWWOz/6aefcvz4cQIDA+3bIBERkSeIQyYP+/fv5+LFiwA0bdoUAG9vb6pUqQLAli1b0i23bds2cy9Carny5ctTokSJNOU2btzIggUL6N69uy43FRERsYFDJg9nz541L3t5eaVZjo2NfaBy58+fZ8SIEZhMJoYOHZp1gYuIiDwFrH4ktyNIvVT0QcuNGjWK69evM3v2bFxcXKyu59ChQ/d1/MxUqFAhy+sUeVTscY7Yk84/eVI87HPPIZOHwoULm5fj4+PNywkJCQD4+PhYVa548eLpljty5AhJSUl07NgRgDt37pjLtW/fnm7duqXbI6FfNCKZ0zki8mjY49yLiIjIcJtDDlv4+/vj4eEBQHh4OABxcXHs3bsXgLp16wIQGBhIYGAgc+fOBaB27drkyJHDotzhw4fNl3emlgNISUnhxo0b3Lhxg8TERPP6W7duWbwWERERSw7Z8+Ds7ExwcDCjRo0iLCyMxo0bc+nSJa5fv07+/Pnp06cPAFFRUQDmyZUFCxakd+/ezJgxg9DQUDZu3MjZs2cxDIOSJUuaexo2btxocbylS5fy7rvvArBq1SpMJtPDaqqIiMhjxyGTB4COHTuSK1cuQkNDiYyMxMXFhYCAAEJCQjJ9kueQIUPMt9U+deoUuXPnJjAwkJCQENzc3B5iC0RERJ5MTsb9zkJ8ykRERFC9enW71F196I92qVfkYYoY1+NRh3BfTo72f9QhiDyQ4qP226XezP7uOeScBxEREXFcSh5ERETEJkoeRERExCZKHkRERMQmSh5ERETEJkoeRERExCZKHkRERMQmSh5ERETEJkoeRERExCZKHkRERMQmSh5ERETEJkoeRERExCZKHkRERMQmSh5ERETEJkoeRERExCZKHkRERMQmSh5ERETEJkoeRERExCZKHkRERMQmSh5ERETEJkoeRERExCZKHkRERMQmSh5ERETEJkoeRERExCZKHkRERMQmSh5ERETEJkoeRERExCY5HnUAmVmzZg2zZs0iMjISV1dXatWqRUhICCVKlMi03Jw5c5g/fz4nT54kT548NGjQgJCQEAoUKADA0aNHCQ0NZe/evZw7d46UlBSKFi3Kyy+/TLdu3ciRw6HfFhERkUfKYXseFi9eTHBwMAcPHqRgwYIkJycTFhZG586dOX/+fIblJk+ezJgxY4iMjKRIkSLcuHGDpUuX0q1bN27evAnA/v37Wbp0KefPn6dYsWI4OTnx999/8+mnnzJ27NiH1UQREZHHkkMmD4mJiUyYMAGAZs2asWHDBtauXYu7uzvx8fHMmDEj3XIXLlxg1qxZAPTq1YuwsDAWLlyIk5MTUVFRLFiwAIBnnnmGSZMm8ccff7B8+XI2btxIkSJFAFixYsVDaKGIiMjjyyGTh/3793Px4kUAmjZtCoC3tzdVqlQBYMuWLemW27ZtG0lJSRblypcvbx7mSC1Xu3ZtgoKCzMMTHh4eVKxYEQBnZ2c7tEhEROTJ4ZDJw9mzZ83LXl5eaZZjY2OztNzRo0fNiUXnzp3vM2oREZGnw2M1M9AwjCwvt3v3bvr378/NmzcJDAykf//+Ge576NCh+zp+ZipUqJDldYo8KvY4R+xJ5588KR72ueeQyUPhwoXNy/Hx8eblhIQEAHx8fKwqV7x48UzLLV++nJEjR5KUlESXLl14//33yZYt484Y/aIRyZzOEZFHwx7nXkRERIbbHHLYwt/fHw8PDwDCw8MBiIuLY+/evQDUrVsXgMDAQAIDA5k7dy5wdy5D6jyG1HKHDx/mxIkTFuUMw2DSpEkMGzaMlJQU3nvvPT744INMEwcRERG5yyF7HpydnQkODmbUqFGEhYXRuHFjLl26xPXr18mfPz99+vQBICoqCsA8ubJgwYL07t2bGTNmEBoaysaNGzl79iyGYVCyZEk6duwIwNq1a/nmm28AcHd3Z82aNaxZs8Z8/EWLFj3M5oqIiDxWHDJ5AOjYsSO5cuUiNDSUyMhIXFxcCAgIICQkBG9v7wzLDRkyBC8vLxYsWMCpU6fInTs3gYGBhISE4ObmBsDt27fN+1+5coV9+/bZvT0iIiJPCifjfmchPmUiIiKoXr26XequPvRHu9Qr8jBFjOvxqEO4LydH+z/qEEQeSPFR++1Sb2Z/9zTILyIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZR8iAiIiI2UfIgIiIiNlHyICIiIjZx6ORhzZo1tG3blkqVKlGjRg0GDRrEiRMn7lluzpw5BAUF4efnR+3atXn33Xe5cOGCxT4XLlzg3XffpXbt2vj5+REUFMSPP/5or6aIiIg8MXI86gAysnjxYkaOHAlA0aJFuXTpEmFhYezevZsVK1ZQsGDBdMtNnjyZr7/+GoCSJUty9uxZli5dyv/+9z+WLVtGrly5uHHjBt26dSMqKgpXV1eKFClCZGQkn3zyCRcvXmTw4MEPrZ0iIiKPG4fseUhMTGTChAkANGvWjA0bNrB27Vrc3d2Jj49nxowZ6Za7cOECs2bNAqBXr16EhYWxcOFCnJyciIqKYsGCBQAsXLiQqKgonJycWLhwIWFhYfTs2ROAmTNnpumlEBERkf/nkMnD/v37uXjxIgBNmzYFwNvbmypVqgCwZcuWdMtt27aNpKQki3Lly5enRIkSFuV+++03AEqUKEH58uUt9k9KSmL79u1Z3SQREZEnhkMOW5w9e9a87OXllWY5NjbW5nLR0dHmcmfOnEmzT4ECBczLqdv/LSIiwuo22OLbThXtUq/Iw2Sv88PuWvzwqCMQeSDnH8G555DJQ0YMw7BbuXvtU7169fs6toiIyJPGIYctChcubF6Oj483LyckJADg4+PzQOWeeeaZNPv8czl1u4iIiKTlkMmDv78/Hh4eAISHhwMQFxfH3r17Aahbty4AgYGBBAYGMnfuXABq165Njhw5LModPnzYfHlnarnU/0+cOMHhw4ct9s+ZMye1a9e2Z/NEREQea07G/Y4F2NnChQsZNWoU8P+Xal67do38+fOzYsUKvL298fX1BWDAgAEMHDgQgIkTJ5qvxki9VPPWrVuULFmSZcuW4ebmxvXr12nXrh3R0dG4urpSuHBhoqOjAejXrx9Dhgx5+A2Wx1ajRo2IiYmhbdu2fPbZZ486HBERu3PIngeAjh07Mm7cOCpUqMC5c+dwcnIiICCA+fPn4+3tnWG5IUOG8N5771G6dGliYmLIlSsXbdq0Yc6cObi5uQHg7u7OnDlzaNu2Lbly5SImJoZSpUrx7rvv8p///OchtVAepu7du+Pr60v37t0fdSgiWW748OH4+vrSqFGjh37s06dP4+vri6+vL0uXLn3ox7fG1KlTzTFK1nDoCZOtWrWiVatWGW4/cuRImnVOTk68+uqrvPrqq5nWXahQIX1LlAwlJibi7Oz8qMMQyXL6bEtWcNieB3nyGYbB/PnzadeuHZUrV6Zq1aq0a9eOPXv2ALBhwwa6dOlC1apV8ff3p1WrVvz0008WV8akfpsYN24co0ePpmbNmtSuXZsxY8Zw584d8z47d+4EYOfOneYyp0+fZunSpebXa9eupV27dvj5+ZnvBbJ792569+5N9erV8fPzo1mzZnz99dfm+4mIPGqNGjVi2bJlAMTExJg/zzt27DAvL1q0iO7du+Pv789PP/0EQFRUFEOGDDHfor9p06bMmjWLlJQUc93fffcdrVu3pkaNGlSsWJFatWoxYMAAoqKiAFi6dCmNGzc27//uu+9a9PD9s8cvNDSUF154gRo1ajB16lRu3rzJ+++/T9WqVWnYsCGLFy+2aNf58+cZMWIEdevWxc/Pj4YNGzJ+/HgSExPN+/yz/nnz5tGoUSOqVq1K3759OX/+vHmfadOmmcs4ei/J48Khex7kyTZmzBjzZNd8+fLh7e3NsWPHiI6O5tSpU7zzzjvA3ftxuLq6cuTIET766CPOnDlDSEiIRV2zZ8/G3d0dFxcX4uLimDNnDiaTiQ4dOlC5cmWOHTvG9evXcXd3p2zZsgBpvn2988475M+f33xVzo4dO+jVqxd37twhb968FClShOjoaCZPnszRo0eZOHGivd8ikXuqUKECN27c4OLFi+TMmZNnn30WgIMHD5r3GT16NLlz56ZYsWJky5aNEydO0KFDB65cuULevHkpXbo0R48eZdy4cZw5c4b3338fuJtsnzx5kmeeeQZvb28iIyNZv349+/fvJzw8HE9PTypUqMChQ4cAKFasGJ6enpQpU8Yixr1793Lw4EHy5s3LhQsXmDZtGuvWrSMhIQE3NzdiY2MZNWoU1apVo0yZMly8eJGOHTsSExODm5sbpUuX5vjx48ycOZNjx47xzTffWNT/v//9j3379vHMM89w48YNNm3axGeffcaECRMoU6YMJ06cIC4uDoDKlSsD4OnpaZ8fyNPCEHkETp06Zfj6+homk8no16+fcevWLcMwDOPSpUvGyZMnjQYNGhgmk8lo166dcevWLSMlJcUYMmSIYTKZjIoVKxoXL140DMMwTCaTYTKZjEaNGhlXrlwxbt26Zbz44ouGyWQy/vOf/5iP161bN8NkMhndunWziGPJkiXmOoKDg43k5GTDMAzjzp07RteuXQ2TyWTUr1/fuHTpkmEYhjFu3Djz/ocPHzYMwzAaNmxomEwmY9iwYfZ+20TSNWzYMMNkMhkNGzY0r/vjjz/Mn9Vu3bqZz7E7d+4Yw4cPN0wmk9G0aVPj6tWrhmEYxooVKwyTyWSUL1/eiI2NNQzDMP7++28jMTHRXOfWrVvNdW7bts0wjLvncuq6JUuWWMSVet5VrFjROHXqlHH9+nWjYsWKhslkMmrVqmVcvnzZOHHihLn8Tz/9ZBiGYUydOtUwmUxGjRo1jLi4OMMwDGP37t3m/Xbv3m1Rf/ny5Y1Dhw4ZhmEY/fv3N0wmk1GnTh1zHFOmTDGXlayhYQt5JPbv328efujZsycuLi7A3R6I1G8iAAEBAbi4uODk5ESLFi2Au7cQT73ENlWjRo3IkycPLi4uFC1aFMDmZ5R0796dbNnunhLZs2dn//79wN1Le/PlywdAy5Ytzfv/9ddfNtUv8qh06tTJfI5lz56dP//8E4Do6GiqV6+Or68vQ4cOBSAlJcW8PTY2lh49elCtWjXKly9vfgYQYP4mb41y5cpRtGhR3NzczN/4q1evTt68eSlWrJh5v9T77aQe/9KlS9StWxdfX1+6dOli3m/fvn0W9ZtMJvOjBlJ7PfSMIvvSsIU8EfLmzWteTr3Xh63+ebtykSdJRp9tDw8P87N//snV1ZVTp07Rv39/kpKScHd3p2LFiiQnJ5uHKP45N+JecufObV5OPT9T1zk5OZm3Gf+6c4CbmxvlypVLU1+ePHksXmfF+S+20bssj4S/vz9OTk4YhsHs2bOpUqUKzs7OXL16lZs3b+Lj40NsbCzr16+nZ8+eODs7s2bNGuDujbxSv2VYy9XVFYAbN25kuM8/f4mlxrhr1y62bNnC5cuXyZcvH6tXrzZv9/PzsykGEXtJ/XzfvHkTwzDSfJbT+2wfO3YMNzc3vvnmG3NvwLVr11i/fj3169cnLCzMPDH4u+++o2rVqqxZs4bg4GCLunLlymVevnnzZpa0x9/fn82bN+Pk5MS4cePMCc6tW7fYtGkTderUsam+1PcH7v4OSL1sX+6fkgd5JIoWLUrXrl2ZO3cuv/zyC3Xr1qVQoUKcPHmSDz74gP/85z+88847/PXXXzRq1AgXFxdiYmKAu8McqXcgtVbp0qX57bff+Ouvv3jppZcoVKgQ3333XaZlBg4cSK9evThz5gxNmjTB09PTfDOxFi1a6JpxcRilS5cG7t6KPzAwkHz58tG/f/8M9+/bty+//PILsbGxNGzYkJIlS3L16lXi4uK4c+cObdu2pWzZsmTPnp3k5GRef/11fHx8zFcw/JOnpyceHh5cunSJ8ePHs3z5clq1avVA91Tp2rUrS5Ys4cyZM7Ro0YJSpUpx+/Ztzpw5Q2JiIhs2bLDobbiX1PcH7g49FihQgAkTJlgMmYhtNOdBHpmRI0fywQcf8Oyzz3Lz5k1Onz5NmTJlKFGiBK1bt+arr76iWrVqXL9+nfPnz+Pr68sHH3yQ5puPNXr16kWdOnVwc3Pj77//tmq+Qs2aNZk9ezYvvvgiKSkpxMTEULJkSQYPHsznn39+P00WsYuXX36ZZs2akSdPHqKjo9m3b1+mwwqlSpVi0aJFtGjRAnd3dyIjI0lKSqJGjRq89957wN25A2PHjqVo0aIkJSXh4eGR7hVGTk5OjBkzhhIlSnD79m3+/PPPDJ98bC1PT08WLlxI+/bt8fT05Pjx41y7dg1/f3+Cg4MtnoJsjYYNG9KhQwc8PDyIiYlh3759WdZL8rRy2NtTi4iIiGNSz4OIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNhEyYOIiIjYRMmDiIiI2ETJg4iIiNgkx6MOQESeTBcuXODbb79l06ZNnDlzhty5c1OiRAlatGhBu3btcHd3f9Qhish9UvIgIlnu9OnTdO7cmdy5czN48GB8fX1xcXHh2LFjLF68GA8PD1566aU05RITE3F2dn4EEYuILTRsISJZ7sMPPyRbtmwsWbKEFi1aULZsWYoVK0bDhg356quvaNmyJQC+vr7MmzePAQMGUKVKFSZNmgTAggULCAgIwM/Pj4CAABYtWmRRv6+vL+vWrbNY16hRI7777juLfebOncsbb7xB5cqVadiwIStWrLBzy0WeDkoeRCRLXbx4kd9//52uXbvi5uaW7j5OTk7m5WnTplG/fn1WrVpFly5dWL9+PR9//DGvvvoqq1atokePHnz00Uds3LjR5limTp1Ko0aNWL58OR06dGDYsGHs37//vtsmIndp2EJEstTJkycxDINSpUpZrK9Xrx5Xr14F4KWXXmL06NEABAUF0b59e/N+Q4cOpVWrVnTr1g2AUqVKceDAAWbOnEmjRo1siiUgIIBOnToB8Oabb7Jjxw5mz57N+PHj77t9IqKeBxF5SObNm8fy5cvx9/cnMTHRvN7Pz89iv+PHj1OtWjWLddWrVycyMtLmY1apUiXN6/upR0QsqedBRLJU8eLFcXJy4vjx4xbrixUrBkCuXLks1v/7dUb+OdTh5OSEYRgW25OSku4nXBG5D+p5EJEslT9/fl544QXmzp3L9evXbS5funRp9uzZY7EuIiKCMmXKmF97enpy/vx58+sLFy5YvE61b9++NK9Lly5tc0wiYknJg4hkuQ8//BDDMGjXrh2rV6/m2LFjREVFsXr1ag4fPky2bBn/6nn99ddZuXIl8+bNIzo6mjlz5rBq1Spef/118z61atVi3rx57N+/n4MHDzJ8+HBcXFzS1BUeHs6iRYuIjo5mxowZbN++nVdffdUubRZ5mmjYQkSyXLFixVi2bBkzZszgyy+/5MyZM+TMmZPSpUvTpUsXunbtmmHZJk2aMHLkSEJDQxk7diw+Pj588MEHFpMlhw0bxogRI+jRowdeXl4MHTo0zTAJwMCBAwkLC2PMmDF4enry6aefUqlSJbu0WeRp4mT8e+BQROQJ4Ovry5dffklgYOCjDkXkiaNhCxEREbGJkgcRERGxiYYtRERExCbqeRARERGbKHkQERERmyh5EBEREZsoeRARERGbKHkQERERmyh5EBEREZv8H2p6M9yyHCuMAAAAAElFTkSuQmCC\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8,6));\n",
    "sns.barplot(x=ab_df['group'], y=ab_df['converted'], ci=False);\n",
    "plt.ylim(0, 0.17);\n",
    "plt.title('Conversion rate by experimental samples', pad=20)\n",
    "plt.xlabel('Group', labelpad=15);\n",
    "plt.ylabel('Converted (proportion)', labelpad=15);"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# 4. Тестирование гипотезы"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Z-statistic: 1.83\n",
      "p-value: 0.067\n",
      "CI 95% for control group: [0.123, 0.142]\n",
      "CI 95% for treatment group: [0.111, 0.129]\n"
     ]
    }
   ],
   "source": [
    "from statsmodels.stats.proportion import proportions_ztest, proportion_confint\n",
    "\n",
    "control_results = ab_df[ab_df['group'] == 'control']['converted']\n",
    "treatment_results = ab_df[ab_df['group'] == 'treatment']['converted']\n",
    "\n",
    "n_con = control_results.count()\n",
    "n_treat = treatment_results.count()\n",
    "successes = [control_results.sum(), treatment_results.sum()]\n",
    "nobs = [n_con, n_treat]\n",
    "\n",
    "z_stat, pval = proportions_ztest(successes, nobs=nobs)\n",
    "(lower_con, lower_treat), (upper_con, upper_treat) = proportion_confint(successes, nobs=nobs, alpha=0.05)\n",
    "\n",
    "print(f'Z-statistic: {z_stat:.2f}')\n",
    "print(f'p-value: {pval:.3f}')\n",
    "print(f'CI 95% for control group: [{lower_con:.3f}, {upper_con:.3f}]')\n",
    "print(f'CI 95% for treatment group: [{lower_treat:.3f}, {upper_treat:.3f}]')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# 5. Интерпретация результатов"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "Полученное p-значение = 0.067 слегка выше выбранной альфы 0.05, поэтому согласно статистическому выводу, мы не можем отклонить нулевую гипотезу H0, что означает наш новый вариант работы сервиса никак статистически значимо не влияет на уровень конверсии и понижение уровня конверсии - это случайность.\n",
    "Доверительный интервал для испытуемой группы [0.11, 0.12] или 11% и 12% соответственно:\n",
    "Он не включает наш baseline\n",
    "Он не достигает нашей цели 15%\n",
    "\n",
    "Ответы на вопросы:\n",
    "1. Лучше оставить старый вариант работы сервиса"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Проверка гипотезы для пользователей разных типов.\n",
    "# User type: New"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "outputs": [
    {
     "data": {
      "text/plain": "(147238, 8)"
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#отбираем новых пользователей как целевую группу\n",
    "df_new_users=df[df['user_type']=='new_user']\n",
    "df_new_users.shape"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "outputs": [
    {
     "data": {
      "text/plain": "                                user_id                   timestamp    group  \\\n0  36c2884d-9846-4553-a37f-fa36ce50cb45  2022-05-04 09:13:11.837377  control   \n1  58b4787c-15a1-4fc9-b1b6-a3b61e68e272  2022-05-22 14:59:41.304326  control   \n2  032e7ad0-0822-43c3-bb67-a96b6a52cc52  2022-05-02 17:33:23.178911  control   \n3  05676ce3-4498-47f3-a3f4-56fd04785318  2022-05-11 14:59:29.033575  control   \n4  b696f64f-f2a4-4878-9f22-c53aa82e092a  2022-05-22 05:59:01.010862  control   \n\n   variants  converted        location age_group user_type  \n0  standard          0  United Kingdom     34-41  new_user  \n1  standard          1  United Kingdom     18-25  new_user  \n2  standard          0  United Kingdom     50-57  new_user  \n3  standard          0  United Kingdom     34-41  new_user  \n4  standard          0  United Kingdom       58+  new_user  ",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>user_id</th>\n      <th>timestamp</th>\n      <th>group</th>\n      <th>variants</th>\n      <th>converted</th>\n      <th>location</th>\n      <th>age_group</th>\n      <th>user_type</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>36c2884d-9846-4553-a37f-fa36ce50cb45</td>\n      <td>2022-05-04 09:13:11.837377</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>34-41</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>58b4787c-15a1-4fc9-b1b6-a3b61e68e272</td>\n      <td>2022-05-22 14:59:41.304326</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>1</td>\n      <td>United Kingdom</td>\n      <td>18-25</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>032e7ad0-0822-43c3-bb67-a96b6a52cc52</td>\n      <td>2022-05-02 17:33:23.178911</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>50-57</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>05676ce3-4498-47f3-a3f4-56fd04785318</td>\n      <td>2022-05-11 14:59:29.033575</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>34-41</td>\n      <td>new_user</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>b696f64f-f2a4-4878-9f22-c53aa82e092a</td>\n      <td>2022-05-22 05:59:01.010862</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>58+</td>\n      <td>new_user</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#среди них случайным образом выбираем control и treatment погруппы\n",
    "control_sample = df_new_users[df_new_users['group'] == 'control'].sample(n = required_n, random_state=42)\n",
    "treatment_sample = df_new_users[df_new_users['group'] == 'treatment'].sample(n = required_n, random_state=42)\n",
    "\n",
    "# Сформируем общую случайную выборку для новых юзеров\n",
    "ab_df_new_users = pd.concat([control_sample, treatment_sample], axis=0)\n",
    "ab_df_new_users.reset_index(drop=True, inplace=True)\n",
    "ab_df_new_users.head()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "outputs": [
    {
     "data": {
      "text/plain": "            CR mean    CR std     CR se\ngroup                                  \ncontrol    0.122034  0.327325  0.004764\ntreatment  0.127331  0.333343  0.004852",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>CR mean</th>\n      <th>CR std</th>\n      <th>CR se</th>\n    </tr>\n    <tr>\n      <th>group</th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>control</th>\n      <td>0.122034</td>\n      <td>0.327325</td>\n      <td>0.004764</td>\n    </tr>\n    <tr>\n      <th>treatment</th>\n      <td>0.127331</td>\n      <td>0.333343</td>\n      <td>0.004852</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conversion_rates = ab_df_new_users.groupby('group')['converted']\n",
    "# Стандартное отклонение для сформированной выборки\n",
    "std_proportion = lambda x: np.std(x, ddof=0)\n",
    "# Стандартная ошибка для сформированной выборки\n",
    "se_proportion = lambda x: stats.sem(x, ddof=0)\n",
    "\n",
    "conversion_rates = conversion_rates.agg([np.mean, std_proportion, se_proportion]).rename(\n",
    "        columns={'mean': 'CR mean', '<lambda_0>': 'CR std', '<lambda_1>': 'CR se'})\n",
    "conversion_rates"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "По выборочным статистикам для новых пользователей два эксперимента +/- показывают себя примерно одинаково. Есть небольшое повышение среднего уровня конверсии для новых пользователей"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 576x432 with 1 Axes>",
      "image/png": "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\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8,6));\n",
    "sns.barplot(x=ab_df_new_users['group'], y=ab_df_new_users['converted'], ci=False);\n",
    "plt.ylim(0, 0.17);\n",
    "plt.title('Conversion rate by new users', pad=20)\n",
    "plt.xlabel('Group', labelpad=15);\n",
    "plt.ylabel('Converted (proportion)', labelpad=15);"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Z-statistic: -0.78\n",
      "p-value: 0.436\n",
      "CI 95% for control group: [0.113, 0.131]\n",
      "CI 95% for treatment group: [0.118, 0.137]\n"
     ]
    }
   ],
   "source": [
    "from statsmodels.stats.proportion import proportions_ztest, proportion_confint\n",
    "\n",
    "control_results = ab_df_new_users[ab_df_new_users['group'] == 'control']['converted']\n",
    "treatment_results = ab_df_new_users[ab_df_new_users['group'] == 'treatment']['converted']\n",
    "\n",
    "n_con = control_results.count()\n",
    "n_treat = treatment_results.count()\n",
    "successes = [control_results.sum(), treatment_results.sum()]\n",
    "nobs = [n_con, n_treat]\n",
    "\n",
    "z_stat, pval = proportions_ztest(successes, nobs=nobs)\n",
    "(lower_con, lower_treat), (upper_con, upper_treat) = proportion_confint(successes, nobs=nobs, alpha=0.05)\n",
    "\n",
    "print(f'Z-statistic: {z_stat:.2f}')\n",
    "print(f'p-value: {pval:.3f}')\n",
    "print(f'CI 95% for control group: [{lower_con:.3f}, {upper_con:.3f}]')\n",
    "print(f'CI 95% for treatment group: [{lower_treat:.3f}, {upper_treat:.3f}]')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "Полученное p-значение = 0.436 значительно выше выбранной альфы 0.05, поэтому согласно статистическому выводу, мы не можем отклонить нулевую гипотезу H0, что означает наш новый вариант работы сервиса никак статистически значимо не влияет на уровень конверсии и повышение уровня конверсии - это случайность. Доверительный интервал для испытуемой группы [0.11, 0.13] или 11% и 13% соответственно: Он включает наш baseline, но не достигает нашей цели 15%\n",
    "\n",
    "Ответ: по группе новых пользователей также не стоит менять старую модель."
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Проверка гипотезы для старых пользователей\n",
    "# User Type: Registered users"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "outputs": [
    {
     "data": {
      "text/plain": "(147240, 8)"
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#отбираем зарегестрированных пользователей как целевую группу\n",
    "df_reg_users=df[df['user_type']=='registered_user']\n",
    "df_reg_users.shape"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "outputs": [
    {
     "data": {
      "text/plain": "                                user_id                   timestamp    group  \\\n0  3960d256-ab06-4282-b1bb-1f84a0ad27f4  2022-05-09 10:12:05.412405  control   \n1  e6d3713e-1f4a-4cab-8186-de2c5bc3c799  2022-05-19 08:49:58.838873  control   \n2  e4270216-3f3d-4f9e-b615-621709a42625  2022-05-20 20:55:27.488474  control   \n3  32802f77-5b12-4a02-ad78-d310496ebaea  2022-05-06 14:03:09.808327  control   \n4  e3ea432a-2313-47b0-977d-d2947061bdde  2022-05-04 03:42:37.266315  control   \n\n   variants  converted        location age_group        user_type  \n0  standard          1  United Kingdom     34-41  registered_user  \n1  standard          0  United Kingdom     26-33  registered_user  \n2  standard          0  United Kingdom     42-49  registered_user  \n3  standard          0  United Kingdom       58+  registered_user  \n4  standard          0  United Kingdom       58+  registered_user  ",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>user_id</th>\n      <th>timestamp</th>\n      <th>group</th>\n      <th>variants</th>\n      <th>converted</th>\n      <th>location</th>\n      <th>age_group</th>\n      <th>user_type</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>3960d256-ab06-4282-b1bb-1f84a0ad27f4</td>\n      <td>2022-05-09 10:12:05.412405</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>1</td>\n      <td>United Kingdom</td>\n      <td>34-41</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>e6d3713e-1f4a-4cab-8186-de2c5bc3c799</td>\n      <td>2022-05-19 08:49:58.838873</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>26-33</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>e4270216-3f3d-4f9e-b615-621709a42625</td>\n      <td>2022-05-20 20:55:27.488474</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>42-49</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>32802f77-5b12-4a02-ad78-d310496ebaea</td>\n      <td>2022-05-06 14:03:09.808327</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>58+</td>\n      <td>registered_user</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>e3ea432a-2313-47b0-977d-d2947061bdde</td>\n      <td>2022-05-04 03:42:37.266315</td>\n      <td>control</td>\n      <td>standard</td>\n      <td>0</td>\n      <td>United Kingdom</td>\n      <td>58+</td>\n      <td>registered_user</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#среди них случайным образом выбираем control и treatment погруппы\n",
    "control_sample = df_reg_users[df_reg_users['group'] == 'control'].sample(n = required_n, random_state=42)\n",
    "treatment_sample = df_reg_users[df_reg_users['group'] == 'treatment'].sample(n = required_n, random_state=42)\n",
    "\n",
    "# Сформируем общую случайную выборку для новых юзеров\n",
    "ab_df_reg_users = pd.concat([control_sample, treatment_sample], axis=0)\n",
    "ab_df_reg_users.reset_index(drop=True, inplace=True)\n",
    "ab_df_reg_users.head()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "outputs": [
    {
     "data": {
      "text/plain": "            CR mean    CR std     CR se\ngroup                                  \ncontrol    0.121398  0.326590  0.004754\ntreatment  0.119280  0.324117  0.004718",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>CR mean</th>\n      <th>CR std</th>\n      <th>CR se</th>\n    </tr>\n    <tr>\n      <th>group</th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>control</th>\n      <td>0.121398</td>\n      <td>0.326590</td>\n      <td>0.004754</td>\n    </tr>\n    <tr>\n      <th>treatment</th>\n      <td>0.119280</td>\n      <td>0.324117</td>\n      <td>0.004718</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ab_df_reg_users.group.value_counts(normalize=True)\n",
    "conversion_rates = ab_df_reg_users.groupby('group')['converted']\n",
    "# Стандартное отклонение для сформированной выборки\n",
    "std_proportion = lambda x: np.std(x, ddof=0)\n",
    "# Стандартная ошибка для сформированной выборки, формулу стандартной ошибки вы знаете из лекций. Забыли - погуглите.\n",
    "se_proportion = lambda x: stats.sem(x, ddof=0)\n",
    "\n",
    "conversion_rates = conversion_rates.agg([np.mean, std_proportion, se_proportion]).rename(\n",
    "        columns={'mean': 'CR mean', '<lambda_0>': 'CR std', '<lambda_1>': 'CR se'})\n",
    "conversion_rates"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "По выборочным статистикам для зарегистрированных пользователей два эксперимента +/- показывают себя примерно одинаково. Наблюдается незначительное понижение уровня конверсии для нового варианта работы сервиса."
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 576x432 with 1 Axes>",
      "image/png": "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\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 6));\n",
    "sns.barplot(x=ab_df_reg_users['group'], y=ab_df_reg_users['converted'], ci=False);\n",
    "plt.ylim(0, 0.17);\n",
    "plt.title('Conversion rate by registered users', pad=20)\n",
    "plt.xlabel('Group', labelpad=15);\n",
    "plt.ylabel('Converted (proportion)', labelpad=15);"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Z-statistic: 0.32\n",
      "p-value: 0.752\n",
      "CI 95% for control group: [0.112, 0.131]\n",
      "CI 95% for treatment group: [0.110, 0.129]\n"
     ]
    }
   ],
   "source": [
    "from statsmodels.stats.proportion import proportions_ztest, proportion_confint\n",
    "\n",
    "control_results = ab_df_reg_users[ab_df_reg_users['group'] == 'control']['converted']\n",
    "treatment_results = ab_df_reg_users[ab_df_reg_users['group'] == 'treatment']['converted']\n",
    "\n",
    "n_con = control_results.count()\n",
    "n_treat = treatment_results.count()\n",
    "successes = [control_results.sum(), treatment_results.sum()]\n",
    "nobs = [n_con, n_treat]\n",
    "\n",
    "z_stat, pval = proportions_ztest(successes, nobs=nobs)\n",
    "(lower_con, lower_treat), (upper_con, upper_treat) = proportion_confint(successes, nobs=nobs, alpha=0.05)\n",
    "\n",
    "print(f'Z-statistic: {z_stat:.2f}')\n",
    "print(f'p-value: {pval:.3f}')\n",
    "print(f'CI 95% for control group: [{lower_con:.3f}, {upper_con:.3f}]')\n",
    "print(f'CI 95% for treatment group: [{lower_treat:.3f}, {upper_treat:.3f}]')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "Полученное p-значение = 0.752 значительно выше выбранной альфы 0.05, поэтому согласно статистическому выводу, мы не можем отклонить нулевую гипотезу H0, что означает наш новый вариант работы сервиса никак статистически значимо не влияет на уровень конверсии и понижение уровня конверсии - это случайность. Доверительный интервал для испытуемой группы [0.11, 0.12] или 11% и 12% соответственно: Он не включает наш baseline и не достигает нашей цели 15%\n",
    "\n",
    "Ответ: по группе зарегистрированных пользователей также не стоит менять старый вариант работы сервиса.\n",
    "\n",
    "2. Итого ни для новых, ни для старых пользователей не стоит менять старый вариант работы сервиса."
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}