analysis_exploratory.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exploratory Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# imports\n",
    "import sys\n",
    "import os\n",
    "import functools\n",
    "import pathlib\n",
    "import glob\n",
    "import collections\n",
    "import itertools\n",
    "import re\n",
    "import random\n",
    "import math\n",
    "try:\n",
    "    import cPickle as pickle\n",
    "except ModuleNotFoundError:\n",
    "    import pickle\n",
    "\n",
    "# from entropy import *\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import pipeline\n",
    "\n",
    "from sklearn import impute\n",
    "from sklearn import datasets\n",
    "from sklearn import svm, linear_model, metrics\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import accuracy_score, classification_report, precision_recall_fscore_support\n",
    "from sklearn.model_selection import KFold, StratifiedKFold, train_test_split\n",
    "from sklearn.preprocessing import StandardScaler, MinMaxScaler\n",
    "\n",
    "import statsmodels.api as sm\n",
    "import statsmodels.formula.api as smf\n",
    "from statsmodels.graphics.regressionplots import abline_plot\n",
    "\n",
    "import scipy\n",
    "from scipy.spatial.distance import cdist\n",
    "from scipy import stats\n",
    "from scipy.stats import pearsonr\n",
    "\n",
    "import shap \n",
    "\n",
    "# visualization libraries\n",
    "%matplotlib inline\n",
    "import matplotlib as mpl\n",
    "from matplotlib.dates import DateFormatter\n",
    "from matplotlib.cbook import boxplot_stats\n",
    "import matplotlib.dates as mdates\n",
    "import matplotlib.transforms as mtrans\n",
    "import seaborn as sns\n",
    "sns.set_style(\"whitegrid\")\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "plt.rcParams.update({'figure.autolayout': True})\n",
    "# plt.rcParams.update({'figure.facecolor': [1.0, 1.0, 1.0, 1.0]})\n",
    "\n",
    "# configure autoreloading of modules\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Another precommit hook test\n",
    "# Read in data\n",
    "master_dataset = pipeline.import_processed_files()\n",
    "master_featureset = pipeline.create_file_dictionary('features')['features']\n",
    "\n",
    "# Obtain baseline features\n",
    "features_bl = master_featureset[pipeline.ALL_USERS_DIR.replace('/', '')]['pre_post']\n",
    "features_bl"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Establish common vars"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Ensure epoch labels align with their correct columns\n",
    "epoch_bins = pipeline.EPOCHS['labels'][1:] + [pipeline.EPOCHS['labels'][0]]\n",
    "epoch_bin_labels = [label.capitalize().replace(\"_\", \" \") for label in epoch_bins]\n",
    "\n",
    "app_launch = master_dataset['app_launch']\n",
    "\n",
    "# Force convert the date column, since pandas seems to struggle with date\n",
    "#   parsing for this project\n",
    "app_launch['date'] = pd.to_datetime(app_launch['date'])\n",
    "\n",
    "# Get list of number of users per app, for convenience\n",
    "apps_nusers = app_launch.groupby('package')['pid'].nunique().reset_index(name=\"num_users\").sort_values(by='num_users', ascending=False)\n",
    "# df.to_csv(\"features/app_usercounts.csv\")\n",
    "apps_nusers\n",
    "\n",
    "# apps = apps_nusers[apps_nusers['num_users'] > 6]\n",
    "apps = list(apps_nusers['package'].unique())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualize Percentage of Missing Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "unused_dfs = ['act', 'loc_semantic_labels']\n",
    "\n",
    "size_all_dfs = 0\n",
    "isna_all_dfs = 0\n",
    "\n",
    "percent_missings_ind = {}\n",
    "\n",
    "for name, df in master_dataset.items():\n",
    "    if name not in unused_dfs:\n",
    "        isna = df.isna().sum().sum()\n",
    "        size = df.size\n",
    "        \n",
    "        isna_all_dfs += isna\n",
    "        size_all_dfs += size\n",
    "        \n",
    "        percent_missings_ind[name] = isna / size * 100\n",
    "        \n",
    "percent_missing = isna_all_dfs / size_all_dfs * 100\n",
    "print(percent_missing)\n",
    "for name, pm in percent_missings_ind.items():\n",
    "    print(name)\n",
    "    print(str(pm))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "master_dataset['blsurvey']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "master_dataset['pre_post']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Confirm nobody with completely missing bl surveys was included in analysis\n",
    "pids = master_dataset['app_launch'].pid.unique()\n",
    "print(pids)\n",
    "master_dataset['blsurvey'][master_dataset['blsurvey']['startdate'].isna()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pids = master_dataset['app_launch'].pid.unique()\n",
    "print(pids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pids_dropped = master_dataset['dropouts'][master_dataset['dropouts']['dropped'] == 1]['pid'].unique()\n",
    "set(pids) - set(pids_dropped)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Demographics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Age\n",
    "TODO - better to get mean age per phil's other paper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = master_dataset['blsurvey']\n",
    "age_df = df[['q13']]\n",
    "age_df.rename(columns={'q13': 'age'}, inplace=True)\n",
    "age_df['age'].mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Race"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dem_df = df[['q15']]\n",
    "dem_df.rename(columns={'q15': 'race'}, inplace=True)\n",
    "np.round(dem_df['race'].value_counts() / len(dem_df) * 100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Importance Graphs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.rcParams.update({'font.size': 20})\n",
    "# RF x FS3 in paper\n",
    "for suffix in ['survey_app_overall_fs_RF_optimized', 'survey_app_mua_fs_XGB_optimized']:\n",
    "\n",
    "    shap_scores = pickle.load(open('feature_importance/shap_' + suffix + '.ob', 'rb'))\n",
    "    if 'RF_' in suffix:\n",
    "        shap_scores = shap_scores[1]\n",
    "        \n",
    "    X_test = pd.read_pickle('feature_importance/X_test_'+ suffix + '.ob')\n",
    "    X_test.columns = [x.replace('_', ' ').capitalize() for x in X_test.columns]\n",
    "    X_test.columns = [x.replace('std', 'SD') for x in X_test.columns]\n",
    "    X_test.columns = [x.replace('Num', 'Number of') for x in X_test.columns]\n",
    "    X_test.columns = [x.replace('min', 'minimum') for x in X_test.columns]\n",
    "    X_test.columns = [x.replace('max', 'maximum') for x in X_test.columns]\n",
    "    X_test.columns = [x.replace(' tob', ' tobacco') for x in X_test.columns]\n",
    "    X_test.columns = [x.replace('Betweenlaunch', 'Between launch') for x in X_test.columns]\n",
    "    X_test.columns = [x.replace('Daysofuse', 'Days of use') for x in X_test.columns]\n",
    "    X_test.columns = [x.replace('Weekofstudy', 'Week of study') for x in X_test.columns]\n",
    "\n",
    "    shap.summary_plot(shap_scores, X_test, show=False, plot_type='bar', max_display=10)\n",
    "    \n",
    "    # Get the current figure and axes objects. from @GarrettCGraham code\n",
    "    fig, ax = plt.gcf(), plt.gca()\n",
    "    \n",
    "    fig.set_size_inches(12.5, 8.5)\n",
    "    \n",
    "    # Modifying main plot parameters\n",
    "    ax.tick_params(labelsize=20)\n",
    "    ax.set_xlabel('mean(|SHAP value|)', fontsize=20)\n",
    "    \n",
    "    plt.savefig('feature_importance/' + suffix + '.png')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Stat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = master_featureset[pipeline.APP_USERS_DIR]['wkly_agg']\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### T-test?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Too sparse for t-test as-is. Let Lee know.\n",
    "\n",
    "df = master_featureset[pipeline.APP_USERS_DIR]['wkly_agg']\n",
    "\n",
    "# Quick look at distributions\n",
    "low_df = df[df['trait_anx_group'] == 'low']\n",
    "high_df = df[df['trait_anx_group'] == 'high']\n",
    "\n",
    "\n",
    "plt.hist(low_df['anx'])\n",
    "plt.hist(high_df['anx'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "scipy.stats.ttest_ind(low_df['frequency'], high_df['frequency'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for group, group_label in pipeline.TRAIT_AFFECT_GROUPS.items():\n",
    "    low_df = df[df[group] == 'low']\n",
    "    high_df = df[df[group] == 'high']\n",
    "\n",
    "    for metric in pipeline.FILENAME_MAPPINGS['metric'].keys():\n",
    "        low_metric = low_df[metric]\n",
    "        high_metric = high_df[metric]\n",
    "        print('Group: ' + group_label + ', Metric: ' + metric)\n",
    "        print('Result: ' + str(scipy.stats.ttest_ind(low_metric, high_metric)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Correlation Heatmaps "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "master_featureset[pipeline.ALL_USERS_DIR]['study_agg']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sns.set(font_scale=3)\n",
    "feats_allmetrics = master_featureset[pipeline.ALL_USERS_DIR]['study_agg']\n",
    "\n",
    "# Add pipeline.TRAIT_AFFECT_SUBSCALES.keys()   (scores) to feats_allmetrics\n",
    "for score in pipeline.TRAIT_AFFECT_SUBSCALES.keys()  :\n",
    "    feats_allmetrics[score] = [pipeline.match_value(features_bl, 'pid', pid, score) for pid in feats_allmetrics['pid']]\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(20, 10)) \n",
    "cbar_ax = fig.add_axes([1.01, .3, .03, .4])\n",
    " \n",
    "fig.suptitle('Trait Affect Correlation with Engagement Metrics', \n",
    "             y=1.05, \n",
    "             fontsize=50)\n",
    "\n",
    "df = feats_allmetrics[list(pipeline.FILENAME_MAPPINGS['metric'].keys()) + \n",
    "                      list(pipeline.TRAIT_AFFECT_SUBSCALES.keys()  )]\n",
    "\n",
    "# Correlate the baseline scores (pipeline.TRAIT_AFFECT_SUBSCALES.keys()  ) with app-specific features\n",
    "#  (e.g., correlate PHQ4 baseline depression with frequency for the \n",
    "#   \"Worrknot\" app)\n",
    "res = df.corr().round(decimals=2)\n",
    "\n",
    "# Identify significant correlations, and create a mask so that we only display those\n",
    "p_values = pipeline.corr_sig(res)\n",
    "mask = np.invert(np.tril(p_values<0.05))\n",
    "\n",
    "# Construct corr plot as seaborn heatmap\n",
    "\n",
    "# if metric == 'daysofuse':\n",
    "#     title = 'Days of Use'\n",
    "\n",
    "\n",
    "sns.heatmap(res, \n",
    "            vmax=1, vmin=-1, \n",
    "            cmap='YlGnBu', \n",
    "            annot=True,\n",
    "            ax=ax,\n",
    "            mask=mask,\n",
    "            cbar=True)\n",
    "\n",
    "# Tidy up the plot\n",
    "\n",
    "xlabels = ax.xaxis.get_majorticklabels()\n",
    "ylabels = ax.yaxis.get_majorticklabels()\n",
    "\n",
    "plt.setp(xlabels, rotation=45,fontsize=34)\n",
    "plt.setp(ylabels, rotation=0)\n",
    "\n",
    "fig.text(0.52, 0, \"Trait Affect Subscale Score\", ha='center')\n",
    "fig.text(-0.01, 0.5, 'Epoch', va='center', rotation='vertical')\n",
    "\n",
    "plt.savefig('results/cor/cor_study.png', bbox_inches = \"tight\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Weekly x Epoch and Study x Epoch Trait Affect x Engagement"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sns.set(font_scale=3)\n",
    "for timediv in ['wkly', 'study']:\n",
    "\n",
    "    # Get all individual epoch/applevel features\n",
    "    feats_allmetrics = master_featureset[timediv + '_epoch_agg']\n",
    "\n",
    "    # Add pipeline.TRAIT_AFFECT_SUBSCALES.keys()   (scores) to feats_allmetrics\n",
    "    for score in pipeline.TRAIT_AFFECT_SUBSCALES.keys()  :\n",
    "        feats_allmetrics[score] = [pipeline.match_value(features_bl, 'pid', pid, score) for pid in feats_allmetrics['pid']]\n",
    "\n",
    "    fig, axs = plt.subplots(1,3, sharex=True, sharey=False, figsize=(70, 25)) \n",
    "    cbar_ax = fig.add_axes([1.01, .3, .03, .4])\n",
    "\n",
    "    title = 'Trait Affect Correlation with'\n",
    "    if timediv == 'wkly':\n",
    "        title = title + ' Weekly'\n",
    "    title = title + ' Metrics by Epoch'\n",
    "    if timediv == 'study':\n",
    "        title = title + ''\n",
    "    fig.suptitle(title, \n",
    "                 y=1.05, \n",
    "                 fontsize=50)\n",
    "\n",
    "    for i, ax in enumerate(axs.flat):\n",
    "        metric = list(pipeline.FILENAME_MAPPINGS['metric'].keys())[i]\n",
    "\n",
    "        # Find the abbreviation for the given metric (e.g. 'frequency' is denoted\n",
    "        #   with 'loy')\n",
    "        metric_abbrev = pipeline.FILENAME_MAPPINGS['metric'][metric]\n",
    "\n",
    "        # Get the subset of features for the current metric\n",
    "        df = feats_allmetrics[[col for col in feats_allmetrics.columns if metric in col] + list(pipeline.TRAIT_AFFECT_SUBSCALES.keys()  )]\n",
    "        \n",
    "        if metric == \"duration\":\n",
    "            df = df[list(df.columns)[0:5] + list(pipeline.TRAIT_AFFECT_SUBSCALES.keys()  )]\n",
    "\n",
    "        # Correlate the baseline scores (pipeline.TRAIT_AFFECT_SUBSCALES.keys()  ) with app-specific features\n",
    "        #  (e.g., correlate PHQ4 baseline depression with frequency for the \n",
    "        #   \"Worrknot\" app)\n",
    "        res = df.corr().round(decimals=2)\n",
    "\n",
    "        # Identify significant correlations, and create a mask so that we only display those\n",
    "        p_values = pipeline.corr_sig(res)\n",
    "        mask = np.invert(np.tril(p_values<0.05))\n",
    "        \n",
    "        # Construct corr plot as seaborn heatmap\n",
    "                        \n",
    "        title =  metric.capitalize()\n",
    "        if metric == 'daysofuse':\n",
    "            title = 'Days of Use'\n",
    "\n",
    "        ax.set_title(title, fontsize='42')\n",
    "        ax.set(xlabel= '(' + pipeline.SUBPLOT_LABELS[i] + ')', \n",
    "               ylabel='')\n",
    "\n",
    "        sns.heatmap(res, \n",
    "                    vmax=1, vmin=-1, \n",
    "                    cmap='YlGnBu', \n",
    "                    annot=True, \n",
    "#                     xticklabels=pipeline.TRAIT_AFFECT_SUBSCALES.values(),\n",
    "#                     yticklabels=epoch_bin_labels,\n",
    "                    ax=ax,\n",
    "                    mask=mask,\n",
    "                    cbar=(i==0),\n",
    "                    cbar_ax=None if i else cbar_ax)\n",
    "\n",
    "        # Tidy up the plot\n",
    "\n",
    "        xlabels = ax.xaxis.get_majorticklabels()\n",
    "        ylabels = ax.yaxis.get_majorticklabels()\n",
    "\n",
    "        plt.setp(xlabels, rotation=45,fontsize=34)\n",
    "        plt.setp(ylabels, rotation=0)\n",
    "\n",
    "    fig.text(0.52, 0, \"Trait Affect Subscale Score\", ha='center')\n",
    "    fig.text(-0.01, 0.5, 'Epoch', va='center', rotation='vertical')\n",
    "\n",
    "    plt.savefig('results/cor/cor_' + timediv + '.png', bbox_inches = \"tight\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Boxplots"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.APP_USERS_DIR]['study_agg'].copy()\n",
    "\n",
    "# Convert duration to minutes, for graphing\n",
    "features['duration'] = features['duration'] / 60.0 \n",
    "\n",
    "# Get only the core engagement vars\n",
    "features = pd.melt(features, \n",
    "                   id_vars=['pid'], \n",
    "                   value_vars=['frequency', 'daysofuse', 'duration'])\n",
    "\n",
    "sns.set(font_scale=1.25)\n",
    "\n",
    "with sns.axes_style(\"whitegrid\"):\n",
    "    fig, axs = plt.subplots(1,3, figsize=(7, 4))\n",
    "    fig.suptitle('Comparison of Metrics', \n",
    "                 y=1.05,\n",
    "                fontsize=20)\n",
    "\n",
    "    for i, ax in enumerate(axs):\n",
    "        metric = list(pipeline.FILENAME_MAPPINGS['metric'].keys())[i]\n",
    "        abbrev = pipeline.FILENAME_MAPPINGS['metric'][metric]  \n",
    "        df = features[features['variable'] == metric]\n",
    "\n",
    "        g = sns.boxplot(x='variable', \n",
    "                        y='value',\n",
    "                        showfliers = False,\n",
    "                        data=df, \n",
    "                        ax=ax)\n",
    "\n",
    "        g.set_xticklabels('')\n",
    "        \n",
    "        title = metric.capitalize()\n",
    "        if metric == \"duration\":\n",
    "            title = title + ' (Minutes)'\n",
    "        if metric == 'daysofuse':\n",
    "            title = 'Days of Use'\n",
    "\n",
    "        ax.set(title=title, xlabel= '(' + pipeline.SUBPLOT_LABELS[i] + ')', ylabel='')\n",
    "    \n",
    "#     plt.savefig('results/box/box_study.png', bbox_inches = \"tight\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Study Trait Affect Engagement Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.APP_USERS_DIR]['study_agg'].copy()\n",
    "sns.set(font_scale=1.25)\n",
    "\n",
    "n_groups = len(pipeline.TRAIT_AFFECT_GROUPS)\n",
    "n_metrics = len(pipeline.FILENAME_MAPPINGS['metric'])\n",
    "\n",
    "with sns.axes_style(\"whitegrid\"):\n",
    "    fig, axs = plt.subplots(n_groups,n_metrics, sharex=True, sharey=False, figsize=(10, 8))\n",
    "    fig.suptitle('Comparison of Engagement Metrics \\nAcross Study Lifetime', \n",
    "                 y=1.15)\n",
    "    \n",
    "    for i in range(0, n_groups):\n",
    "        \n",
    "        group = list(pipeline.TRAIT_AFFECT_GROUPS.keys())[i]\n",
    "        group_label = pipeline.TRAIT_AFFECT_GROUPS[group]\n",
    "        group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]\n",
    "    \n",
    "        if i == 0:\n",
    "            y = 0.65\n",
    "        else:\n",
    "            y = 0.15\n",
    "        fig.text(-.05, y, group_label.replace(' Group', ''), weight='bold', ha='center', rotation = 90)\n",
    "        \n",
    "        for j in range(0, n_metrics):\n",
    "            ax = axs[i,j]\n",
    "            metric = list(pipeline.FILENAME_MAPPINGS['metric'].keys())[j]\n",
    "            abbrev = pipeline.FILENAME_MAPPINGS['metric'][metric]            \n",
    "       \n",
    "            g = None\n",
    "            outliers = None\n",
    "            if metric == 'duration':\n",
    "                features[metric] = features[metric] / 60.0\n",
    "\n",
    "            g = sns.boxplot(x=features[group], \n",
    "                            y=features[metric],\n",
    "                            hue=features[group], \n",
    "                            palette={'high': 'darkorange', 'low': 'royalblue'},\n",
    "                            showfliers = False,\n",
    "                            dodge=False,\n",
    "                            data=features, \n",
    "                            ax=ax)\n",
    "            \n",
    "            outliers_low = [y for stat in boxplot_stats(features[features[group] == 'low'][metric]) for y in stat['fliers']]\n",
    "            outliers_high = [y for stat in boxplot_stats(features[features[group] == 'high'][metric]) for y in stat['fliers']]\n",
    "            print(outliers_low)\n",
    "            print(outliers_high)\n",
    "    \n",
    "            g.set_xticklabels(\"\")\n",
    "\n",
    "            title = metric.capitalize()\n",
    "            if metric == \"duration\":\n",
    "                title = title + ' (Minutes)'\n",
    "            if metric == 'daysofuse':\n",
    "                title = 'Days of Use'\n",
    "\n",
    "            if i == 0:            \n",
    "                ax.set(title=title, xlabel= '(' + pipeline.SUBPLOT_LABELS[j] + ')', ylabel='')\n",
    "            else:\n",
    "                ax.set(title=\"\", xlabel= '(' + pipeline.SUBPLOT_LABELS[j+3] + ')', ylabel='')\n",
    "            \n",
    "            if i == 0 and j == 2:\n",
    "                g.legend(title='Trait Group',\n",
    "                         bbox_to_anchor=(1.05, 1), ncol=1)\n",
    "                legend = g.get_legend()\n",
    "#                 plt.setp(legend.get_texts(), fontsize='28') # for legend text\n",
    "#                 plt.setp(legend.get_title(), fontsize='30') # for legend title\n",
    "#                 plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "            else: \n",
    "                g.legend_.remove()\n",
    "\n",
    "#     plt.savefig('results/box/box_study.png', bbox_inches = \"tight\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.APP_USERS_DIR]['study_agg'].copy()\n",
    "sns.set(font_scale=1.25)\n",
    "\n",
    "n_groups = len(pipeline.TRAIT_AFFECT_GROUPS)\n",
    "n_metrics = len(pipeline.FILENAME_MAPPINGS['metric'])\n",
    "\n",
    "with sns.axes_style(\"whitegrid\"):\n",
    "    fig, axs = plt.subplots(n_groups,n_metrics, sharex=True, sharey=False, figsize=(10, 7))\n",
    "    fig.suptitle('Comparison of Engagement Metrics (Mean and Std) \\nAcross Study Lifetime', \n",
    "                 y=1.15)\n",
    "    \n",
    "    for i in range(0, n_groups):\n",
    "        \n",
    "        group = list(pipeline.TRAIT_AFFECT_GROUPS.keys())[i]\n",
    "        group_label = pipeline.TRAIT_AFFECT_GROUPS[group]\n",
    "        group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]\n",
    "    \n",
    "        if i == 0:\n",
    "            y = 0.65\n",
    "        else:\n",
    "            y = 0.15\n",
    "        fig.text(-.05, y, group_label.replace(' Group', ''), weight='bold', ha='center', rotation = 90)\n",
    "        \n",
    "        for j in range(0, n_metrics):\n",
    "            ax = axs[i,j]\n",
    "            metric = list(pipeline.FILENAME_MAPPINGS['metric'].keys())[j]\n",
    "            abbrev = pipeline.FILENAME_MAPPINGS['metric'][metric]            \n",
    "       \n",
    "            g = None\n",
    "            outliers = None\n",
    "            if metric == 'duration':\n",
    "                features[metric] = features[metric] / 60.0\n",
    "\n",
    "            g = sns.pointplot(x=features[group], \n",
    "                y=features[metric],\n",
    "                hue=features[group], \n",
    "                ci = 68,\n",
    "                palette={'high': 'darkorange', 'low': 'royalblue'},\n",
    "                dodge=False,\n",
    "                connect = False,\n",
    "                ax=ax)\n",
    "\n",
    "    \n",
    "            g.set_xticklabels(\"\")\n",
    "\n",
    "            title = metric.capitalize()\n",
    "            if metric == \"duration\":\n",
    "                title = title + ' (Minutes)'\n",
    "            if metric == 'daysofuse':\n",
    "                title = 'Days of Use'\n",
    "\n",
    "            if i == 0:            \n",
    "                ax.set(title=title, xlabel= '(' + pipeline.SUBPLOT_LABELS[j] + ')', ylabel='')\n",
    "            else:\n",
    "                ax.set(title=\"\", xlabel= '(' + pipeline.SUBPLOT_LABELS[j+3] + ')', ylabel='')\n",
    "            \n",
    "            if i == 0 and j == 2:\n",
    "                g.legend(title='Trait Group',\n",
    "                         bbox_to_anchor=(1.05, 1), ncol=1)\n",
    "                legend = g.get_legend()\n",
    "#                 plt.setp(legend.get_texts(), fontsize='28') # for legend text\n",
    "#                 plt.setp(legend.get_title(), fontsize='30') # for legend title\n",
    "#                 plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "            else: \n",
    "                g.legend_.remove()\n",
    "\n",
    "    plt.savefig('results/box/mean_std_study.png', bbox_inches = \"tight\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sns.set(font_scale=3)\n",
    "feats_allmetrics = master_featureset[pipeline.ALL_USERS_DIR]['study_agg']\n",
    "\n",
    "# Add pipeline.TRAIT_AFFECT_SUBSCALES.keys()   (scores) to feats_allmetrics\n",
    "for score in pipeline.TRAIT_AFFECT_SUBSCALES.keys()  :\n",
    "    feats_allmetrics[score] = [pipeline.match_value(features_bl, 'pid', pid, score) for pid in feats_allmetrics['pid']]\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(20, 10)) \n",
    "cbar_ax = fig.add_axes([1.01, .3, .03, .4])\n",
    " \n",
    "fig.suptitle('Trait Affect Correlation with Engagement Metrics', \n",
    "             y=1.05, \n",
    "             fontsize=50)\n",
    "\n",
    "df = feats_allmetrics[list(pipeline.FILENAME_MAPPINGS['metric'].keys()) + \n",
    "                      list(pipeline.TRAIT_AFFECT_SUBSCALES.keys()  )]\n",
    "\n",
    "# Correlate the baseline scores (pipeline.TRAIT_AFFECT_SUBSCALES.keys()  ) with app-specific features\n",
    "#  (e.g., correlate PHQ4 baseline depression with frequency for the \n",
    "#   \"Worrknot\" app)\n",
    "res = df.corr().round(decimals=2)\n",
    "\n",
    "# Identify significant correlations, and create a mask so that we only display those\n",
    "p_values = pipeline.corr_sig(res)\n",
    "mask = np.invert(np.tril(p_values<0.05))\n",
    "\n",
    "# Construct corr plot as seaborn heatmap\n",
    "\n",
    "# if metric == 'daysofuse':\n",
    "#     title = 'Days of Use'\n",
    "\n",
    "\n",
    "sns.heatmap(res, \n",
    "            vmax=1, vmin=-1, \n",
    "            cmap='YlGnBu', \n",
    "            annot=True,\n",
    "            ax=ax,\n",
    "            mask=mask,\n",
    "            cbar=True)\n",
    "\n",
    "# Tidy up the plot\n",
    "\n",
    "xlabels = ax.xaxis.get_majorticklabels()\n",
    "ylabels = ax.yaxis.get_majorticklabels()\n",
    "\n",
    "plt.setp(xlabels, rotation=45,fontsize=34)\n",
    "plt.setp(ylabels, rotation=0)\n",
    "\n",
    "fig.text(0.52, 0, \"Trait Affect Subscale Score\", ha='center')\n",
    "fig.text(-0.01, 0.5, 'Epoch', va='center', rotation='vertical')\n",
    "\n",
    "plt.savefig('results/cor/cor_study.png', bbox_inches = \"tight\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Study x Epoch Trait Affect Engagement Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Get Study x Epoch Longform df\n",
    "\n",
    "features = master_featureset[pipeline.APP_USERS_DIR]['study_epoch_agg_lf']\n",
    "sns.set(font_scale=3)\n",
    "\n",
    "for group, group_label in pipeline.TRAIT_AFFECT_GROUPS.items():\n",
    "\n",
    "    group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]    \n",
    "    \n",
    "    with sns.axes_style(\"whitegrid\"):\n",
    "        fig, axs = plt.subplots(1,3, sharex=True, sharey=False, figsize=(25, 9))\n",
    "        fig.suptitle(group_label + ' Comparison of Metrics by Epoch',  \n",
    "                     y=1.05, \n",
    "                     fontsize=50)\n",
    "\n",
    "        for i, ax in enumerate(axs):\n",
    "            metric = list(pipeline.FILENAME_MAPPINGS['metric'].keys())[i]\n",
    "            abbrev = pipeline.FILENAME_MAPPINGS['metric'][metric]            \n",
    "       \n",
    "            g = None\n",
    "            if metric == 'duration':\n",
    "                g = sns.boxplot(x=features['epoch'], \n",
    "                                y=features[metric] / 60.0, # Convert to minutes, for display, \n",
    "                                hue=features[group], \n",
    "                                palette={'high': 'sandybrown', 'low': 'cornflowerblue'},\n",
    "                                showfliers = False,\n",
    "                                data=features, \n",
    "                                ax=ax)\n",
    "            else:\n",
    "                g = sns.boxplot(x=features['epoch'], \n",
    "                                y=features[metric],\n",
    "                                hue=features[group], \n",
    "                                palette={'high': 'sandybrown', 'low': 'cornflowerblue'},\n",
    "                                showfliers = False,\n",
    "                                data=features, \n",
    "                                ax=ax)\n",
    "    \n",
    "            g.set_xticklabels(epoch_bin_labels, rotation=45)\n",
    "\n",
    "            title = metric.capitalize()\n",
    "            if metric == \"duration\":\n",
    "                title = title + ' (Minutes)'\n",
    "\n",
    "            if metric == 'daysofuse':\n",
    "                title = 'Days of Use'\n",
    "\n",
    "            ax.set(title=title, xlabel= '(' + pipeline.SUBPLOT_LABELS[i] + ')', ylabel='')\n",
    "\n",
    "            \n",
    "            if metric == \"duration\":\n",
    "                g.legend(title='Trait Group',\n",
    "                         bbox_to_anchor=(1.05, 1), ncol=1).texts[0].set_text('')\n",
    "                legend = g.get_legend()\n",
    "                plt.setp(legend.get_texts(), fontsize='28') # for legend text\n",
    "                plt.setp(legend.get_title(), fontsize='30') # for legend title\n",
    "                plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "            else: \n",
    "                g.legend_.remove()\n",
    "\n",
    "    fig.text(0.52, 0, 'Epoch', ha='center')\n",
    "\n",
    "    #                 ax.legend(title=\"Baseline Group\")\n",
    "\n",
    "#             plt.setp(ax.get_legend().get_lines(), linewidth='4')\n",
    "\n",
    "#         plt.gcf().subplots_adjust(bottom=0.15)\n",
    "    \n",
    "    plt.savefig('results/box/box_study_epoch_' + group_abbrev + '.png', bbox_inches = \"tight\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Weekly x Epoch Trait Affect Engagement Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.APP_USERS_DIR]['wkly_epoch_agg_lf']\n",
    "sns.set(font_scale=3)\n",
    "\n",
    "for group, group_label in pipeline.TRAIT_AFFECT_GROUPS.items():\n",
    "\n",
    "    group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]    \n",
    "    \n",
    "    with sns.axes_style(\"whitegrid\"):\n",
    "        fig, axs = plt.subplots(1,3, sharex=True, sharey=False, figsize=(25, 9))\n",
    "        fig.suptitle(group_label + ' Comparison of Weekly Metrics by Epoch', \n",
    "                     y=1.05, \n",
    "                     fontsize=50)\n",
    "\n",
    "        for i, ax in enumerate(axs):\n",
    "            metric = list(pipeline.FILENAME_MAPPINGS['metric'].keys())[i]\n",
    "            abbrev = pipeline.FILENAME_MAPPINGS['metric'][metric]            \n",
    "       \n",
    "            g = None\n",
    "            if metric == 'duration':\n",
    "                g = sns.boxplot(x=features['epoch'], \n",
    "                                y=features[metric] / 60.0, # Convert to minutes, for display, \n",
    "                                hue=features[group], \n",
    "                                palette={'high': 'sandybrown', 'low': 'cornflowerblue'},\n",
    "                                showfliers = False,\n",
    "                                data=features, \n",
    "                                ax=ax)\n",
    "            else:\n",
    "                g = sns.boxplot(x=features['epoch'], \n",
    "                                y=features[metric],\n",
    "                                hue=features[group], \n",
    "                                palette={'high': 'sandybrown', 'low': 'cornflowerblue'},\n",
    "                                showfliers = False,\n",
    "                                data=features, \n",
    "                                ax=ax)\n",
    "    \n",
    "            g.set_xticklabels(epoch_bin_labels, rotation=45)\n",
    "\n",
    "            title = metric.capitalize()\n",
    "            if metric == \"duration\":\n",
    "                title = title + ' (Minutes)'\n",
    "\n",
    "            if metric == 'daysofuse':\n",
    "                title = 'Days of Use'\n",
    "\n",
    "            ax.set(title=title, xlabel= '(' + pipeline.SUBPLOT_LABELS[i] + ')', ylabel='')\n",
    "            \n",
    "            if metric == \"duration\":\n",
    "                g.legend(title='Trait Group',\n",
    "                         bbox_to_anchor=(1.05, 1), ncol=1)\n",
    "                legend = g.get_legend()\n",
    "                plt.setp(legend.get_texts(), fontsize='28') # for legend text\n",
    "                plt.setp(legend.get_title(), fontsize='30') # for legend title\n",
    "                plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "            else: \n",
    "                g.legend_.remove()\n",
    "\n",
    "\n",
    "    fig.text(0.52, 0, 'Epoch', ha='center')\n",
    "\n",
    "    #                 ax.legend(title=\"Baseline Group\")\n",
    "\n",
    "#             plt.setp(ax.get_legend().get_lines(), linewidth='4')\n",
    "\n",
    "#         plt.gcf().subplots_adjust(bottom=0.15)\n",
    "    \n",
    "    plt.savefig('results/box/box_wkly_epoch_' + group_abbrev + '.png', bbox_inches = \"tight\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Line Plots"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.APP_USERS_DIR]['wkly_agg'].copy()\n",
    "\n",
    "# Convert duration to minutes, for graphing\n",
    "features['duration'] = features['duration'] / 60.0 \n",
    "\n",
    "# Get only the core engagement vars\n",
    "features = pd.melt(features, \n",
    "                   id_vars=['pid','weekofstudy'], \n",
    "                   value_vars=['frequency', 'daysofuse', 'duration'])\n",
    "features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.APP_USERS_DIR]['wkly_agg'].copy()\n",
    "\n",
    "# Convert duration to minutes, for graphing\n",
    "features['duration'] = features['duration'] / 60.0 \n",
    "\n",
    "# Get only the core engagement vars\n",
    "features = pd.melt(features, \n",
    "                   id_vars=['pid','weekofstudy'], \n",
    "                   value_vars=['frequency', 'daysofuse', 'duration'])\n",
    "\n",
    "sns.set(font_scale=1.25)\n",
    "\n",
    "with sns.axes_style(\"whitegrid\"):\n",
    "    fig, axs= plt.subplots(1,3, figsize=(12, 4))\n",
    "    fig.suptitle('Comparison of Metrics', \n",
    "                 y=1.05,\n",
    "                fontsize=20)\n",
    "\n",
    "    for i, ax in enumerate(axs):\n",
    "        metric = list(pipeline.FILENAME_MAPPINGS['metric'].keys())[i]\n",
    "        abbrev = pipeline.FILENAME_MAPPINGS['metric'][metric]  \n",
    "        df = features[features['variable'] == metric]\n",
    "\n",
    "        g = sns.lineplot(x='weekofstudy', \n",
    "                        y='value',\n",
    "                        data=df, \n",
    "                        ci=68,\n",
    "                        ax=ax)\n",
    "        \n",
    "        title = metric.capitalize()\n",
    "        if metric == \"duration\":\n",
    "            title = title + ' (Minutes)'\n",
    "        if metric == 'daysofuse':\n",
    "            title = 'Days of Use'\n",
    "\n",
    "        ax.set(title=title, xlabel= '(' + pipeline.SUBPLOT_LABELS[i] + ')', ylabel='')\n",
    "    fig.text(0.5, 0, 'Week of Study', ha='center')\n",
    "    plt.savefig('results/line/line_study.png', bbox_inches = \"tight\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Weekly trait affect group comparison of engagement metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = master_featureset[pipeline.APP_USERS_DIR]['wkly_agg']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset['app_users_only']['wkly_agg'].copy()\n",
    "features['duration'] = features['duration'] / 60.0 \n",
    "\n",
    "ylims = {\n",
    "    'frequency': [0, 55],\n",
    "    'daysofuse': [0, 7.5],\n",
    "    'duration': [0, 35]\n",
    "}\n",
    "\n",
    "\n",
    "n_groups = len(pipeline.TRAIT_AFFECT_GROUPS)\n",
    "n_metrics = len(pipeline.FILENAME_MAPPINGS['metric'])\n",
    "\n",
    "sns.set(font_scale=1.5)\n",
    "\n",
    "with sns.axes_style(\"whitegrid\"):\n",
    "    fig, axs = plt.subplots(n_groups,n_metrics, sharex=True, sharey=False, figsize=(16, 8))\n",
    "#     fig.suptitle(group_label+ ' Comparison of Weekly Engagement Metrics', \n",
    "#                  y=1.15)\n",
    "    \n",
    "    for i in range(0, n_groups):\n",
    "        group = list(pipeline.TRAIT_AFFECT_GROUPS.keys())[i]\n",
    "        group_label = pipeline.TRAIT_AFFECT_GROUPS[group]\n",
    "        group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]\n",
    "        \n",
    "        df2 = features.groupby([group])['pid'].nunique().reset_index(name='n_participants')\n",
    "\n",
    "        \n",
    "#       Addition of random undersampling\n",
    "#         subgroup = df2.iloc[df2['n_participants'].argmax(), :][group]\n",
    "#         pids = list(df[df[group]==subgroup]['pid'].unique())\n",
    "#         n_to_exclude = df2['n_participants'].argmax() - df2['n_participants'].argmin()\n",
    "#         pids_to_exclude = random.sample(pids, n_to_exclude)\n",
    "#         features = features[~features['pid'].isin(pids_to_exclude)]\n",
    "        \n",
    "        \n",
    "        if i == 0:\n",
    "            y = 0.6\n",
    "        else:\n",
    "            y = 0.1\n",
    "            \n",
    "        # Correction from Phil: Trait -> Baseline (needs to be done throughout)\n",
    "        new_label = group_label.replace('Trait', 'Baseline').replace(' Group', '')\n",
    "        \n",
    "        # Hack for copyediting\n",
    "        new_label = new_label.replace('Anxiety', 'anxiety').replace('Depression', 'depression')\n",
    "        fig.text(-.01, y, new_label, weight='bold', ha='center', rotation = 90)\n",
    "        \n",
    "        for j in range(0, n_metrics):\n",
    "            ax = axs[i,j]\n",
    "            metric = list(pipeline.FILENAME_MAPPINGS['metric'].keys())[j]\n",
    "            abbrev = pipeline.FILENAME_MAPPINGS['metric'][metric]            \n",
    "       \n",
    "            g = sns.lineplot(x='weekofstudy', \n",
    "                             y= metric,\n",
    "                             hue=group,\n",
    "                             palette={'high': 'darkorange', 'low': 'royalblue'},\n",
    "                             ci=68,\n",
    "                             data=features, \n",
    "                             ax=ax)\n",
    "        \n",
    "            ax.set_ylim(ylims[metric])\n",
    "\n",
    "            title = metric.capitalize()\n",
    "            if metric == \"duration\":\n",
    "                title = title + ' (minutes)'\n",
    "            if metric == 'daysofuse':\n",
    "                title = 'Days of use'\n",
    "                        \n",
    "            if i > 0:          \n",
    "                title = ''\n",
    "            \n",
    "            ax.set(title=title, adjustable='datalim', xlim=(0,8),\n",
    "                   xlabel= '', ylabel='')\n",
    "            \n",
    "            if i == 0 and j == 2:\n",
    "                g.legend(title='Baseline Group',\n",
    "                         bbox_to_anchor=(1.05, 1), ncol=1).texts[0].set_text('')\n",
    "                legend = g.get_legend()\n",
    "#                 plt.setp(legend.get_texts(), fontsize='28') # for legend text\n",
    "#                 plt.setp(legend.get_title(), fontsize='30') # for legend title\n",
    "                plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "            else: \n",
    "                g.legend_.remove()\n",
    "            \n",
    "            fig.text(0.165, -0.05, '(' + pipeline.SUBPLOT_LABELS[3].upper() + ')', ha='center')\n",
    "            fig.text(0.475, -0.05, '(' + pipeline.SUBPLOT_LABELS[4].upper() + ')', ha='center')\n",
    "            fig.text(0.75, -0.05, '(' + pipeline.SUBPLOT_LABELS[5].upper() + ')', ha='center')\n",
    "            \n",
    "            fig.text(0.165, 0.49, '(' + pipeline.SUBPLOT_LABELS[0].upper() + ')', ha='center')\n",
    "            fig.text(0.475, 0.49, '(' + pipeline.SUBPLOT_LABELS[1].upper() + ')', ha='center')\n",
    "            fig.text(0.75,  0.49, '(' + pipeline.SUBPLOT_LABELS[2].upper() + ')', ha='center')\n",
    "            \n",
    "            \n",
    "    fig.text(0.475, 0, 'Week of study', ha='center')\n",
    "    plt.savefig('results/line/line_eng_68ci_nounder.png', bbox_inches = \"tight\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Weekly trait affect group comparison of Weekly affect change"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.APP_USERS_DIR]['wkly_agg'].copy()\n",
    "sns.set(font_scale=3)\n",
    "for group, group_label in pipeline.TRAIT_AFFECT_GROUPS.items():\n",
    "    \n",
    "    group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]    \n",
    "\n",
    "    with sns.axes_style(\"whitegrid\"):\n",
    "        fig, axs = plt.subplots(1,2, sharex=True, sharey=False, figsize=(20, 8)) \n",
    "\n",
    "        \n",
    "        fig.suptitle(group_label + ' Comparison of Weekly Affect', \n",
    "                     x=0.5,\n",
    "                     y=1.05, \n",
    "                     fontsize=50)\n",
    "        \n",
    "        for i, ax in enumerate(axs.flat):\n",
    "            wkly_affect = list(pipeline.WKLY_AFFECT.keys())[i]\n",
    "            title = pipeline.WKLY_AFFECT[wkly_affect]\n",
    "            \n",
    "            group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]\n",
    "            \n",
    "            g = sns.lineplot(x='weekofstudy', \n",
    "                         y= wkly_affect,\n",
    "                         hue=group,\n",
    "                         palette={'high': 'sandybrown', 'low': 'cornflowerblue'},\n",
    "                         data=features, \n",
    "                         err_style=\"bars\",\n",
    "                         ci=68,\n",
    "                         ax=ax)\n",
    "\n",
    "            wkly_affect = list(pipeline.WKLY_AFFECT.keys())[i]\n",
    "            \n",
    "            ax.set(adjustable='datalim',\n",
    "                   xlim=(0,8),\n",
    "                   ylim=(0,5.5),\n",
    "                   xlabel= '(' + pipeline.SUBPLOT_LABELS[i] + ')', \n",
    "                   ylabel = '',\n",
    "                   title=title)\n",
    "            \n",
    "#             ax.set_title(metric.capitalize(), fontsize='42')\n",
    "            \n",
    "            if wkly_affect == 'dep':\n",
    "                g.legend(title='Trait Group',\n",
    "                         bbox_to_anchor=(1, 1.05), ncol=1).texts[0].set_text('')\n",
    "                legend = g.get_legend()\n",
    "                plt.setp(legend.get_texts(), fontsize='20') # for legend text\n",
    "                plt.setp(legend.get_title(), fontsize='22') # for legend title\n",
    "                plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "            else: \n",
    "                g.legend_.remove()\n",
    "        \n",
    "            for line in ax.lines:\n",
    "                plt.setp(ax.lines,linewidth=4) \n",
    "            \n",
    "                        \n",
    "            for subgroup in ['low', 'high']:\n",
    "                x = features[features[group]==subgroup]['weekofstudy']\n",
    "                y = features[features[group]==subgroup][wkly_affect]\n",
    "                idx = np.isfinite(x) & np.isfinite(y)\n",
    "                z = np.polyfit(x[idx], y[idx], 1)\n",
    "                p = np.poly1d(z)\n",
    "                if subgroup == 'low':\n",
    "                    c = 'cornflowerblue'\n",
    "                else:\n",
    "                    c = 'sandybrown'\n",
    "                ax.plot(x, p(x), c=c)\n",
    "                \n",
    "       \n",
    "        fig.text(0.44, 0, 'Week of Study', ha='center')\n",
    "        plt.savefig('results/line/line_' + group_abbrev + '_wklyaffect.png', bbox_inches = \"tight\")\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.APP_USERS_DIR]['wkly_agg'].copy()\n",
    "sns.set(font_scale=1.5)\n",
    "\n",
    "with sns.axes_style(\"whitegrid\"):\n",
    "    fig, axs = plt.subplots(1, 2, sharex=True, sharey=False, figsize=(12, 5)) \n",
    "    \n",
    "    fig.suptitle(group_label + ' Comparison of Weekly Affect', \n",
    "                 x=0.5,\n",
    "                 y=1.05, \n",
    "                 fontsize=20)\n",
    "    \n",
    "    for i, ax in enumerate(axs.flat):\n",
    "        group = list(pipeline.TRAIT_AFFECT_GROUPS.keys())[i]\n",
    "        print(group)\n",
    "        group_label = pipeline.TRAIT_AFFECT_GROUPS[group]\n",
    "        group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]\n",
    "    \n",
    "        if 'anx' in group:\n",
    "            wkly_affect = 'anx'\n",
    "        else:\n",
    "            wkly_affect = 'dep'\n",
    "\n",
    "        g = sns.lineplot(x='weekofstudy', \n",
    "                     y= wkly_affect,\n",
    "                     hue=group,\n",
    "                     palette={'high': 'sandybrown', 'low': 'cornflowerblue'},\n",
    "                     data=features, \n",
    "                     err_style=\"bars\",\n",
    "                     ci=68,\n",
    "                     ax=ax)\n",
    "\n",
    "        ax.set(adjustable='datalim',\n",
    "               xlim=(0,8),\n",
    "               ylim=(0,5.5),\n",
    "               title = '',\n",
    "               xlabel= '', \n",
    "               ylabel = pipeline.WKLY_AFFECT[wkly_affect] + ' Score')\n",
    "\n",
    "    #             ax.set_title(metric.capitalize(), fontsize='42')\n",
    "        if wkly_affect == 'dep':\n",
    "            g.legend(title='Trait Group',\n",
    "                     bbox_to_anchor=(1, 1.05), ncol=1).texts[0].set_text('')\n",
    "            legend = g.get_legend()\n",
    "            plt.setp(legend.get_texts(), fontsize='14') # for legend text\n",
    "            plt.setp(legend.get_title(), fontsize='16') # for legend title\n",
    "            plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "        else:\n",
    "            g.legend().remove()\n",
    "\n",
    "        for line in ax.lines:\n",
    "            plt.setp(ax.lines,linewidth=3) \n",
    "\n",
    "\n",
    "        for subgroup in ['low', 'high']:\n",
    "            x = features[features[group]==subgroup]['weekofstudy']\n",
    "            y = features[features[group]==subgroup][wkly_affect]\n",
    "            idx = np.isfinite(x) & np.isfinite(y)\n",
    "            z = np.polyfit(x[idx], y[idx], 1)\n",
    "            p = np.poly1d(z)\n",
    "            if subgroup == 'low':\n",
    "                c = 'cornflowerblue'\n",
    "            else:\n",
    "                c = 'sandybrown'\n",
    "            ax.plot(x, p(x), c=c)\n",
    "                \n",
    "fig.text(0.249, -0.10, '(' + pipeline.SUBPLOT_LABELS[0] + ')', ha='center')\n",
    "fig.text(0.675, -0.10, '(' + pipeline.SUBPLOT_LABELS[1] + ')', ha='center')\n",
    "            \n",
    "fig.text(0.455, 0, 'Week of Study', ha='center')\n",
    "plt.savefig('results/line/line_wklyaffect.png', bbox_inches = \"tight\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pipeline.FILENAME_MAPPINGS['group']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = master_featureset[pipeline.ALL_USERS_DIR]['wkly_trait_state']\n",
    "\n",
    "# Indicator (step) function showing net group change in affect by week\n",
    "# For instance, in the high trait anxious group, let's say we have three people with affect score changes +1 point,\n",
    "#  +3 points, and -2 points between any two given weeks. 1 + 3 - 2 = 2, \n",
    "#  so the indicator function is set to 1 indicating a net positive change between the two weeks.\n",
    "\n",
    "sns.set(font_scale=1.5)\n",
    "    \n",
    "for group, group_label in pipeline.TRAIT_AFFECT_GROUPS.items():\n",
    "    group_abbrev = pipeline.FILENAME_MAPPINGS['group'][group]    \n",
    "    with sns.axes_style(\"whitegrid\"):\n",
    "        fig, axs = plt.subplots(2,2, sharex=True, sharey=True, figsize=(10, 6))\n",
    "        fig.suptitle(group_label + ' Comparison of Change in Weekly Affect', \n",
    "                     x=0.5,\n",
    "                     y=1.05, \n",
    "                     fontsize=20)\n",
    "\n",
    "        for j in range(0,2):\n",
    "            affect = list(pipeline.WKLY_AFFECT.keys())[j]\n",
    "            print(affect)\n",
    "            affect_change = affect + '_change'\n",
    "            features[affect_change] = np.nan\n",
    "            features[affect_change] = features[affect] - features[affect].shift(1)\n",
    "\n",
    "            df = features.groupby([group, 'weekofstudy'])[affect_change].sum().reset_index()\n",
    "            df[affect_change] = df[affect_change].apply(lambda x: 1 if x > 0 else x)\n",
    "            df[affect_change] = df[affect_change].apply(lambda x: -1 if x < 0  else x)\n",
    "\n",
    "            high_df = df[df[group] == 'high']\n",
    "            low_df = df[df[group] == 'low']\n",
    "            \n",
    "            ax = axs[0, j]\n",
    "            ax.set(ylabel='Low Group')\n",
    "\n",
    "            g = sns.lineplot(x='weekofstudy', \n",
    "                             y=affect_change,\n",
    "                             hue=group,\n",
    "                             palette={'low': 'cornflowerblue'},\n",
    "                             drawstyle='steps-pre',\n",
    "                             data=low_df,\n",
    "                             ax=ax)\n",
    "            \n",
    "#             if j == 1:\n",
    "#                 g.legend(title='Trait Group',\n",
    "#                          bbox_to_anchor=(1.05, 2), ncol=1).texts[0].set_text('')\n",
    "#                 legend = g.get_legend()\n",
    "#                 plt.setp(legend.get_texts(), fontsize='12') # for legend text\n",
    "#                 plt.setp(legend.get_title(), fontsize='14') # for legend title\n",
    "#                 plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "#             else: \n",
    "#                 g.legend_.remove()\n",
    "            \n",
    "            g.legend_.remove()\n",
    "\n",
    "            ax = axs[1, j]\n",
    "            ax.set(ylabel='High Group')\n",
    "\n",
    "            g = sns.lineplot(x='weekofstudy', \n",
    "                             y=affect_change,\n",
    "                             hue=group,\n",
    "                             palette={'high': 'sandybrown'},\n",
    "                             drawstyle='steps-pre',\n",
    "                             data=high_df,\n",
    "                             ax=ax)\n",
    "            \n",
    "            g.legend_.remove()\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "            axs[0,j].set(title='Weekly ' + affect.capitalize())\n",
    "\n",
    "#             if j == 1:\n",
    "#                 g.legend(title='Trait Group',\n",
    "#                          bbox_to_anchor=(1.05, 2), ncol=1).texts[0].set_text('')\n",
    "#                 legend = g.get_legend()\n",
    "#                 plt.setp(legend.get_texts(), fontsize='12') # for legend text\n",
    "#                 plt.setp(legend.get_title(), fontsize='14') # for legend title\n",
    "#                 plt.setp(legend.get_lines()[1:], linewidth='6')\n",
    "#             else: \n",
    "#                 g.legend_.remove()\n",
    "        for i, ax in enumerate(axs.flat):\n",
    "            print(i)\n",
    "            ax.set(adjustable='datalim',\n",
    "                   xlim=(0,8),\n",
    "                   xlabel= '(' + pipeline.SUBPLOT_LABELS[i] + ')')\n",
    "            for line in ax.lines:\n",
    "                plt.setp(ax.lines,linewidth=3) \n",
    "\n",
    "fig.text(0.5, 0, 'Week of Study', ha='center')\n",
    "            \n",
    "#     plt.savefig('results/line/line_' + group_abbrev + '_affect_indicator.png', bbox_inches = \"tight\")\n",
    "plt.show()\n",
    "        \n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "int-eng",
   "language": "python",
   "name": "int-eng"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}