example_ps_match.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## A complete example of ps-matching code"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We first read the data and quickly explore it using tableone package"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<frozen importlib._bootstrap>:219: RuntimeWarning: numpy.ufunc size changed, may indicate binary incompatibility. Expected 192 from C header, got 216 from PyObject\n",
      "<frozen importlib._bootstrap>:219: RuntimeWarning: numpy.ufunc size changed, may indicate binary incompatibility. Expected 192 from C header, got 216 from PyObject\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from tableone import TableOne\n",
    "from typing import List\n",
    "\n",
    "from pandas import Series\n",
    "from src.plots import ScorePlotter\n",
    "from src.prepare_data import PrepData\n",
    "from src.propensity_matching import PScorer\n",
    "from src.find_similarities import ObsMatcher"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "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>target</th>\n",
       "      <th>acquirer</th>\n",
       "      <th>order_total_spent</th>\n",
       "      <th>credit_card_type</th>\n",
       "      <th>credit_card_company</th>\n",
       "      <th>model_score_category</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>166518</th>\n",
       "      <td>1</td>\n",
       "      <td>acquirer_A</td>\n",
       "      <td>246</td>\n",
       "      <td>DEBIT</td>\n",
       "      <td>company_2</td>\n",
       "      <td>intermediate_score</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>132508</th>\n",
       "      <td>1</td>\n",
       "      <td>acquirer_B</td>\n",
       "      <td>198</td>\n",
       "      <td>CREDIT</td>\n",
       "      <td>company_1</td>\n",
       "      <td>intermediate_score</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>137852</th>\n",
       "      <td>0</td>\n",
       "      <td>acquirer_A</td>\n",
       "      <td>49</td>\n",
       "      <td>DEBIT</td>\n",
       "      <td>company_2</td>\n",
       "      <td>low_score</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>255871</th>\n",
       "      <td>1</td>\n",
       "      <td>acquirer_A</td>\n",
       "      <td>139</td>\n",
       "      <td>CREDIT</td>\n",
       "      <td>company_2</td>\n",
       "      <td>intermediate_score</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>229889</th>\n",
       "      <td>1</td>\n",
       "      <td>acquirer_A</td>\n",
       "      <td>111</td>\n",
       "      <td>DEBIT</td>\n",
       "      <td>company_1</td>\n",
       "      <td>intermediate_score</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        target    acquirer  order_total_spent credit_card_type credit_card_company model_score_category\n",
       "id                                                                                                     \n",
       "166518       1  acquirer_A                246            DEBIT           company_2   intermediate_score\n",
       "132508       1  acquirer_B                198           CREDIT           company_1   intermediate_score\n",
       "137852       0  acquirer_A                 49            DEBIT           company_2            low_score\n",
       "255871       1  acquirer_A                139           CREDIT           company_2   intermediate_score\n",
       "229889       1  acquirer_A                111            DEBIT           company_1   intermediate_score"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv('data/df.csv', index_col='id')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th colspan=\"5\" halign=\"left\">Grouped by acquirer</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>Missing</th>\n",
       "      <th>Overall</th>\n",
       "      <th>acquirer_A</th>\n",
       "      <th>acquirer_B</th>\n",
       "      <th>SMD (acquirer_A,acquirer_B)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>n</th>\n",
       "      <th></th>\n",
       "      <td></td>\n",
       "      <td>49995</td>\n",
       "      <td>38708</td>\n",
       "      <td>11287</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total Spent, mean (SD)</th>\n",
       "      <th></th>\n",
       "      <td>0</td>\n",
       "      <td>131.8 (63.7)</td>\n",
       "      <td>128.4 (63.9)</td>\n",
       "      <td>143.6 (61.5)</td>\n",
       "      <td>0.242</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">credit_card_type, n (%)</th>\n",
       "      <th>CREDIT</th>\n",
       "      <td>0</td>\n",
       "      <td>22998 (46.0)</td>\n",
       "      <td>16743 (43.3)</td>\n",
       "      <td>6255 (55.4)</td>\n",
       "      <td>0.245</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DEBIT</th>\n",
       "      <td></td>\n",
       "      <td>26997 (54.0)</td>\n",
       "      <td>21965 (56.7)</td>\n",
       "      <td>5032 (44.6)</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">credit_card_company, n (%)</th>\n",
       "      <th>company_1</th>\n",
       "      <td>0</td>\n",
       "      <td>28435 (56.9)</td>\n",
       "      <td>24081 (62.2)</td>\n",
       "      <td>4354 (38.6)</td>\n",
       "      <td>0.487</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>company_2</th>\n",
       "      <td></td>\n",
       "      <td>21560 (43.1)</td>\n",
       "      <td>14627 (37.8)</td>\n",
       "      <td>6933 (61.4)</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">model_score_category, n (%)</th>\n",
       "      <th>high_score</th>\n",
       "      <td>0</td>\n",
       "      <td>12586 (25.2)</td>\n",
       "      <td>10725 (27.7)</td>\n",
       "      <td>1861 (16.5)</td>\n",
       "      <td>0.385</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>intermediate_score</th>\n",
       "      <td></td>\n",
       "      <td>24909 (49.8)</td>\n",
       "      <td>19658 (50.8)</td>\n",
       "      <td>5251 (46.5)</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>low_score</th>\n",
       "      <td></td>\n",
       "      <td>12500 (25.0)</td>\n",
       "      <td>8325 (21.5)</td>\n",
       "      <td>4175 (37.0)</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>authorization, n (%)</th>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>41901 (83.8)</td>\n",
       "      <td>32957 (85.1)</td>\n",
       "      <td>8944 (79.2)</td>\n",
       "      <td>0.155</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div><br />"
      ],
      "text/plain": [
       "                                               Grouped by acquirer                                                                      \n",
       "                                                           Missing       Overall    acquirer_A    acquirer_B SMD (acquirer_A,acquirer_B)\n",
       "n                                                                          49995         38708         11287                            \n",
       "Total Spent, mean (SD)                                           0  131.8 (63.7)  128.4 (63.9)  143.6 (61.5)                       0.242\n",
       "credit_card_type, n (%)     CREDIT                               0  22998 (46.0)  16743 (43.3)   6255 (55.4)                       0.245\n",
       "                            DEBIT                                   26997 (54.0)  21965 (56.7)   5032 (44.6)                            \n",
       "credit_card_company, n (%)  company_1                            0  28435 (56.9)  24081 (62.2)   4354 (38.6)                       0.487\n",
       "                            company_2                               21560 (43.1)  14627 (37.8)   6933 (61.4)                            \n",
       "model_score_category, n (%) high_score                           0  12586 (25.2)  10725 (27.7)   1861 (16.5)                       0.385\n",
       "                            intermediate_score                      24909 (49.8)  19658 (50.8)   5251 (46.5)                            \n",
       "                            low_score                               12500 (25.0)   8325 (21.5)   4175 (37.0)                            \n",
       "authorization, n (%)        1                                    0  41901 (83.8)  32957 (85.1)   8944 (79.2)                       0.155"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Columns to summarize\n",
    "columns = ['order_total_spent', 'credit_card_type', 'credit_card_company',\n",
    "           'model_score_category', 'target']\n",
    "\n",
    "# Columns containing categorical variables\n",
    "categorical = ['credit_card_type', 'credit_card_company', 'model_score_category', 'target']\n",
    "\n",
    "# Stratification variable - The Intervention\n",
    "groupby = ['acquirer']\n",
    "\n",
    "# Rename columns\n",
    "labels={'order_total_spent': 'Total Spent', 'target': 'authorization'}\n",
    "\n",
    "# Keep level 1 in target column\n",
    "limit = {\"target\": 1}\n",
    "\n",
    "table1 = TableOne(df, columns, categorical, groupby, pval=False, smd=True,\n",
    "                  htest_name=True, limit=limit, rename=labels)\n",
    "table1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we set the relevant variables and initiate the PrepData class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "PS_GROUP = 'acquirer'     # set the group variable (treatment/control)\n",
    "TARGET = 'target'         # set the target variable, the outcome of interest\n",
    "FILE_PATH = 'data/df.csv' # dataframe contains all dependant and independent variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loaded data with 49995 observations\n",
      "False    0.774237\n",
      "True     0.225763\n",
      "Name: acquirer, dtype: float64\n",
      "The minority class contains 11287 observations\n",
      "False    0.838104\n",
      "True     0.161896\n",
      "Name: target, dtype: float64\n",
      "The minority class contains 8094 observations\n"
     ]
    }
   ],
   "source": [
    "data = PrepData(FILE_PATH, group=PS_GROUP, target=TARGET,  index_col=\"id\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We initiate the PScorer and estimate the Propensity Score, currently only logostic regression is availble"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "id\n",
       "166518    0.350140\n",
       "132508   -0.209982\n",
       "137852    0.581685\n",
       "255871    0.582715\n",
       "229889   -0.783610\n",
       "            ...   \n",
       "137660   -0.869582\n",
       "140041    1.225633\n",
       "191360   -0.173064\n",
       "247552   -0.829147\n",
       "292476   -0.514117\n",
       "Length: 49995, dtype: float64"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scorer = PScorer()\n",
    "scorer.fit(data.input, data.group_label)\n",
    "ps_scores: Series = scorer.predict(data.input)\n",
    "ps_scores"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Using ROC curve and AUC is a useful way to assess the model fit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "ScorePlotter.plot_roc_curve(ps_scores, data.group_label)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We initate the ObsMatcher class, set the matching ratio and the caliper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 28/11287 [00:00<00:41, 272.88it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "starting matching process, this might take a time\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 11287/11287 [00:24<00:00, 465.61it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "please note:the matched dataset contains 10659 observations from minority class\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "matcher = ObsMatcher(n_matches=1, caliper=0.001)\n",
    "matched_index: List[int] = matcher.match_scores(ps_scores, data.group_label)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Using SMD plot, before and after matching, we can asses the matching quality"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    " ScorePlotter.plot_smd_comparison(\n",
    "        data=data.input,\n",
    "        matched_index=matched_index,\n",
    "        treatment=data.group_label\n",
    "        )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We now can run again the tableone and assess the causal effect on the matched population"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "acquirer_A    10659\n",
       "acquirer_B    10659\n",
       "Name: acquirer, dtype: int64"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_matched = df[df.index.isin(matched_index)]\n",
    "df_matched.acquirer.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th colspan=\"5\" halign=\"left\">Grouped by acquirer</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>Missing</th>\n",
       "      <th>Overall</th>\n",
       "      <th>acquirer_A</th>\n",
       "      <th>acquirer_B</th>\n",
       "      <th>SMD (acquirer_A,acquirer_B)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>n</th>\n",
       "      <th></th>\n",
       "      <td></td>\n",
       "      <td>21318</td>\n",
       "      <td>10659</td>\n",
       "      <td>10659</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Total Spent, mean (SD)</th>\n",
       "      <th></th>\n",
       "      <td>0</td>\n",
       "      <td>141.0 (61.9)</td>\n",
       "      <td>141.1 (62.9)</td>\n",
       "      <td>141.0 (60.9)</td>\n",
       "      <td>-0.001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">credit_card_type, n (%)</th>\n",
       "      <th>CREDIT</th>\n",
       "      <td>0</td>\n",
       "      <td>11775 (55.2)</td>\n",
       "      <td>5935 (55.7)</td>\n",
       "      <td>5840 (54.8)</td>\n",
       "      <td>0.018</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DEBIT</th>\n",
       "      <td></td>\n",
       "      <td>9543 (44.8)</td>\n",
       "      <td>4724 (44.3)</td>\n",
       "      <td>4819 (45.2)</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">credit_card_company, n (%)</th>\n",
       "      <th>company_1</th>\n",
       "      <td>0</td>\n",
       "      <td>8574 (40.2)</td>\n",
       "      <td>4275 (40.1)</td>\n",
       "      <td>4299 (40.3)</td>\n",
       "      <td>0.005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>company_2</th>\n",
       "      <td></td>\n",
       "      <td>12744 (59.8)</td>\n",
       "      <td>6384 (59.9)</td>\n",
       "      <td>6360 (59.7)</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">model_score_category, n (%)</th>\n",
       "      <th>high_score</th>\n",
       "      <td>0</td>\n",
       "      <td>3711 (17.4)</td>\n",
       "      <td>1855 (17.4)</td>\n",
       "      <td>1856 (17.4)</td>\n",
       "      <td>0.023</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>intermediate_score</th>\n",
       "      <td></td>\n",
       "      <td>10471 (49.1)</td>\n",
       "      <td>5291 (49.6)</td>\n",
       "      <td>5180 (48.6)</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>low_score</th>\n",
       "      <td></td>\n",
       "      <td>7136 (33.5)</td>\n",
       "      <td>3513 (33.0)</td>\n",
       "      <td>3623 (34.0)</td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>authorization, n (%)</th>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>17610 (82.6)</td>\n",
       "      <td>8965 (84.1)</td>\n",
       "      <td>8645 (81.1)</td>\n",
       "      <td>0.079</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div><br />"
      ],
      "text/plain": [
       "                                               Grouped by acquirer                                                                      \n",
       "                                                           Missing       Overall    acquirer_A    acquirer_B SMD (acquirer_A,acquirer_B)\n",
       "n                                                                          21318         10659         10659                            \n",
       "Total Spent, mean (SD)                                           0  141.0 (61.9)  141.1 (62.9)  141.0 (60.9)                      -0.001\n",
       "credit_card_type, n (%)     CREDIT                               0  11775 (55.2)   5935 (55.7)   5840 (54.8)                       0.018\n",
       "                            DEBIT                                    9543 (44.8)   4724 (44.3)   4819 (45.2)                            \n",
       "credit_card_company, n (%)  company_1                            0   8574 (40.2)   4275 (40.1)   4299 (40.3)                       0.005\n",
       "                            company_2                               12744 (59.8)   6384 (59.9)   6360 (59.7)                            \n",
       "model_score_category, n (%) high_score                           0   3711 (17.4)   1855 (17.4)   1856 (17.4)                       0.023\n",
       "                            intermediate_score                      10471 (49.1)   5291 (49.6)   5180 (48.6)                            \n",
       "                            low_score                                7136 (33.5)   3513 (33.0)   3623 (34.0)                            \n",
       "authorization, n (%)        1                                    0  17610 (82.6)   8965 (84.1)   8645 (81.1)                       0.079"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "table2 = TableOne(df_matched, columns, categorical, groupby, pval=False, smd=True,\n",
    "                  htest_name=True, limit=limit, rename=labels)\n",
    "table2"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}