{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Showing uncertainty\n",
"> Uncertainty occurs everywhere in data science, but it's frequently left out of visualizations where it should be included. Here, we review what a confidence interval is and how to visualize them for both single estimates and continuous functions. Additionally, we discuss the bootstrap resampling technique for assessing uncertainty and how to visualize it properly. This is the Summary of lecture \"Improving Your Data Visualizations in Python\", via datacamp.\n",
"\n",
"- toc: true \n",
"- badges: true\n",
"- comments: true\n",
"- author: Chanseok Kang\n",
"- categories: [Python, Datacamp, Visualization]\n",
"- image: images/so2_compare.png"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import pandas as pd\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"import seaborn as sns\n",
"\n",
"plt.rcParams['figure.figsize'] = (10, 5)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Point estimate intervals\n",
"- When is uncertainty important?\n",
" - Estimates from sample\n",
" - Average of a subset\n",
" - Linear model coefficients\n",
"- Why is uncertainty important?\n",
" - Helps inform confidence in estimate\n",
" - Neccessary for decision making\n",
" - Acknowledges limitations of data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Basic confidence intervals\n",
"You are a data scientist for a fireworks manufacturer in Des Moines, Iowa. You need to make a case to the city that your company's large fireworks show has not caused any harm to the city's air. To do this, you look at the average levels for pollutants in the week after the fourth of July and how they compare to readings taken after your last show. By showing confidence intervals around the averages, you can make a case that the recent readings were well within the normal range."
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# Construct CI bounds for averages\n",
"average_ests['lower'] = average_ests['mean'] - 1.96 * average_ests['std_err']\n",
"average_ests['upper'] = average_ests['mean'] + 1.96 * average_ests['std_err']\n",
"\n",
"# Setup a grid of plots, with non-shared x axes limits\n",
"g = sns.FacetGrid(average_ests, row='pollutant', sharex=False, aspect=2);\n",
"\n",
"# Plot CI for average estimate\n",
"g.map(plt.hlines, 'y', 'lower', 'upper');\n",
"\n",
"# Plot observed values for comparison and remove axes labels\n",
"g.map(plt.scatter, 'seen', 'y', color='orangered').set_ylabels('').set_xlabels('');"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This simple visualization shows that all the observed values fall well within the confidence intervals for all the pollutants except for $O_3$."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Annotating confidence intervals\n",
"Your data science work with pollution data is legendary, and you are now weighing job offers in both Cincinnati, Ohio and Indianapolis, Indiana. You want to see if the SO2 levels are significantly different in the two cities, and more specifically, which city has lower levels. To test this, you decide to look at the differences in the cities' SO2 values (Indianapolis' - Cincinnati's) over multiple years.\n",
"\n",
"Instead of just displaying a p-value for a significant difference between the cities, you decide to look at the 95% confidence intervals (columns `lower` and `upper`) of the differences. This allows you to see the magnitude of the differences along with any trends over the years."
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# Set start and ends according to intervals\n",
"# Make intervals thicker\n",
"plt.hlines(y='year', xmin='lower', xmax='upper', \n",
" linewidth=5, color='steelblue', alpha=0.7,\n",
" data=diffs_by_year);\n",
"\n",
"# Point estimates\n",
"plt.plot('mean', 'year', 'k|', data=diffs_by_year);\n",
"\n",
"# Add a 'null' reference line at 0 and color orangered\n",
"plt.axvline(x=0, color='orangered', linestyle='--');\n",
"\n",
"# Set descriptive axis labels and title\n",
"plt.xlabel('95% CI');\n",
"plt.title('Avg SO2 differences between Cincinnati and Indianapolis');"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"By looking at the confidence intervals you can see that the difference flipped from generally positive (more pollution in Cincinnati) in 2013 to negative (more pollution in Indianapolis) in 2014 and 2015. Given that every year's confidence interval contains the null value of zero, no P-Value would be significant, and a plot that only showed significance would have been entirely hidden this trend."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Confidence bands"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Making a confidence band\n",
"Vandenberg Air Force Base is often used as a location to launch rockets into space. You have a theory that a recent increase in the pace of rocket launches could be harming the air quality in the surrounding region. To explore this, you plotted a 25-day rolling average line of the measurements of atmospheric $NO_2$. To help decide if any pattern observed is random-noise or not, you decide to add a 99% confidence band around your rolling mean. Adding a confidence band to a trend line can help shed light on the stability of the trend seen. This can either increase or decrease the confidence in the discovered trend.\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# Draw 99% interval bands for average NO2\n",
"vandenberg_NO2['lower'] = vandenberg_NO2['mean'] - 2.58 * vandenberg_NO2['std_err']\n",
"vandenberg_NO2['upper'] = vandenberg_NO2['mean'] + 2.58 * vandenberg_NO2['std_err']\n",
"\n",
"# Plot mean estimate as a white semi-transparent line\n",
"plt.plot('day', 'mean', data=vandenberg_NO2, color='white', alpha=0.4);\n",
"\n",
"# Fill between the upper and lower confidence band values\n",
"plt.fill_between(x='day', y1='lower', y2='upper', data=vandenberg_NO2);"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This plot shows that the middle of the year's $NO_2$ values are not only lower than the beginning and end of the year but also are less noisy. If just the moving average line were plotted, then this potentially interesting observation would be completely missed. (Can you think of what may cause reduced variance at the lower values of the pollutant?)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Separating a lot of bands\n",
"It is relatively simple to plot a bunch of trend lines on top of each other for rapid and precise comparisons. Unfortunately, if you need to add uncertainty bands around those lines, the plot becomes very difficult to read. Figuring out whether a line corresponds to the top of one class' band or the bottom of another's can be hard due to band overlap. Luckily in Seaborn, it's not difficult to break up the overlapping bands into separate faceted plots.\n",
"\n",
"To see this, explore trends in SO2 levels for a few cities in the eastern half of the US. If you plot the trends and their confidence bands on a single plot - it's a mess. To fix, use Seaborn's `FacetGrid()` function to spread out the confidence intervals to multiple panes to ease your inspection."
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# setup a grid of plots with columns divided by location\n",
"g = sns.FacetGrid(eastern_SO2, col='city', col_wrap=2);\n",
"\n",
"# Map interval plots to each cities data with coral colored ribbons\n",
"g.map(plt.fill_between, 'day', 'lower', 'upper', color='coral');\n",
"\n",
"# Map overlaid mean plots with white line\n",
"g.map(plt.plot, 'day', 'mean', color='white');"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"By separating each band into its own plot you can investigate each city with ease. Here, you see that Des Moines and Houston on average have lower SO2 values for the entire year than the two cities in the Midwest. Cincinnati has a high and variable peak near the beginning of the year but is generally more stable and lower than Indianapolis."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Cleaning up bands for overlaps\n",
"You are working for the city of Denver, Colorado and want to run an ad campaign about how much cleaner Denver's air is than Long Beach, California's air. To investigate this claim, you will compare the SO2 levels of both cities for the year 2014. Since you are solely interested in how the cities compare, you want to keep the bands on the same plot. To make the bands easier to compare, decrease the opacity of the confidence bands and set a clear legend."
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"for city, color in [('Denver', '#66c2a5'), ('Long Beach', '#fc8d62')]:\n",
" # Filter data to desired city\n",
" city_data = SO2_compare[SO2_compare.city == city]\n",
" \n",
" # Set city interval color to desired and lower opacity\n",
" plt.fill_between(x='day', y1='lower', y2='upper', data=city_data, color=color, alpha=0.4);\n",
" \n",
" # Draw a faint mean line for reference and give a label for legend\n",
" plt.plot('day', 'mean', data=city_data, label=city, color=color, alpha=0.25);\n",
" \n",
"plt.legend();"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"From these two curves you can see that during the first half of the year Long Beach generally has a higher average SO2 value than Denver, in the middle of the year they are very close, and at the end of the year Denver seems to have higher averages. However, by showing the confidence intervals, you can see however that almost none of the year shows a statistically meaningful difference in average values between the two cities."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Beyond 95%\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 90, 95, and 99% intervals\n",
"You are a data scientist for an outdoor adventure company in Fairbanks, Alaska. Recently, customers have been having issues with SO2 pollution, leading to costly cancellations. The company has sensors for CO, NO2, and O3 but not SO2 levels.\n",
"\n",
"You've built a model that predicts SO2 values based on the values of pollutants with sensors (loaded as `pollution_model`, a `statsmodels` object). You want to investigate which pollutant's value has the largest effect on your model's SO2 prediction. This will help you know which pollutant's values to pay most attention to when planning outdoor tours. To maximize the amount of information in your report, show multiple levels of uncertainty for the model estimates."
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [],
"source": [
"from statsmodels.formula.api import ols"
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {},
"outputs": [],
"source": [
"pollution = pd.read_csv('./dataset/pollution_wide.csv')\n",
"pollution = pollution.query(\"city == 'Fairbanks' & year == 2014 & month == 11\")"
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {},
"outputs": [],
"source": [
"pollution_model = ols(formula='SO2 ~ CO + NO2 + O3 + day', data=pollution)\n",
"res = pollution_model.fit()"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
""
]
},
"execution_count": 16,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAnAAAAEvCAYAAADFI/wcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nO3df5xVdb3v8ddnYABRKJQfQhwkAxWZYsQ5Rhmil0OZnjQl5HDVJEmupLfMLH/gL5TT5Zj0UIs0uJkcI1Ljhx6jDuWhIDtEI0EiZIipECQ01UUaUZTv/WM2PAaYkcFmZu8vvJ6Px36w9nd911qftb8PZt7zXXvtHSklJEmSlI+yYhcgSZKkA2OAkyRJyowBTpIkKTMGOEmSpMwY4CRJkjJjgJMkScpM22IX0Jq6du2a+vbtW+wyJEmS9uupp576U0qpW0PrDqkA17dvX6qrq4tdhiRJ0n5FxIuNrfMSqiRJUmYMcJIkSZkxwEmSJGXmkHoPnCRJh4odO3awYcMGtm/fXuxStB8dOnSgd+/elJeXN3kbA5wkSQehDRs20KlTJ/r27UtEFLscNSKlRE1NDRs2bODd7353k7fzEqokHeRWD/t4sUtQEWzfvp2jjjrK8FbiIoKjjjrqgGdKDXCSJB2kDG95eDvjVLIBLiJujYhril2HJOWu30PTi12CDlFt2rShsrKSiooKRo0aRW1t7Vv2/+AHP7jffd51112N7uf000/n+OOPp7KyksrKSj7xiU80up8VK1awYMGC3c8fe+wxpkyZst/jN8Vb1dhcSjbASZKaR7ujuxe7BB2iDjvsMFasWMGqVato164d991331v2/8UvfrHffe4vHM2aNYsVK1awYsUKvv/97zfab+8Ad84553Ddddft9/hNccgFuIiYGBHPRsRPgOMLbZdFxK8iYmVEzImIjhHRKSJ+HxHlhT6dI+KFXc8lSVJpGTp0KM899xwAX/3qV6moqKCiooK77rprd58jjjgCgJ/+9KecfvrpfOITn+CEE07gwgsvJKXEPffcw8aNGznjjDM444wzmnzsRx55hIqKCgYNGsRpp53G66+/zs0338xDDz1EZWUlDz30EA888ABXXnklAGPHjmXChAmcccYZHHvssfzsZz/j0ksvZcCAAYwdO3b3fidMmEBVVRUDBw7klltuAWiwxoULF/KBD3yAwYMHM2rUKLZt2/Z3vZZA3d0PpfAATgaeBjoCnYHngGuAo+r1mQz878Lyt4GPF5bHA1Mb2e94oBqo7tOnT5Ik6VCwevXqYpeQDj/88JRSSjt27EjnnHNO+sY3vpGqq6tTRUVF2rZtW3rllVfSiSeemJYvX75H/0WLFqXOnTun9evXpzfffDMNGTIkLVmyJKWU0jHHHJO2bNnS4PGGDRuWjjvuuDRo0KA0aNCgdM0116SUUqqoqEgbNmxIKaX0l7/8JaWU0re//e10xRVX7N62/vNLLrkkjR49Ou3cuTPNnz8/derUKf3mN79Jb775Zho8eHD69a9/nVJKqaamJqWU0htvvJGGDRuWVq5cuU+NW7ZsSUOHDk3btm1LKaU0ZcqUNGnSpH1qb2i8gOrUSG4qpRm4ocC8lFJtSmkr8FihvSIilkTE08CFwMBC+/8FPlVY/hR1gW4fKaXpKaWqlFJVt24Nfh+sJElqAa+++iqVlZVUVVXRp08fxo0bx89//nPOO+88Dj/8cI444gjOP/98lixZss+2p5xyCr1796asrIzKykpeeOGFJh2z/iXUr3zlKwCceuqpjB07lhkzZvDmm282aT8f+9jHiAje+9730qNHD9773vdSVlbGwIEDd9fy8MMPM3jwYE466SSeeeYZVq9evc9+li5dyurVqzn11FOprKxk5syZvPhio19x2mSl9jlwqYG2B6ibaVsZEWOB0wFSSk9GRN+IGAa0SSmtarUqJUnSfu16D1x9dRNL+9e+ffvdy23atOGNN95423Xcd999/PKXv+QHP/gBlZWV+9T0VscvKyvbo5aysjLeeOMNfv/733PnnXfyq1/9ii5dujB27NgGPwokpcSIESOYPXv2266/IaU0A7cYOC8iDouITsDHCu2dgE2F97dduNc2/w7MppHZN0kSvP7HzcUuQdrttNNOY/78+dTW1vK3v/2NefPmMXTo0CZv36lTJ1555ZUDOua6det4//vfz2233UbXrl1Zv37929pPfVu3buXwww/nHe94By+//DI//OEPG6xxyJAhPPnkk7vf/1dbW8vvfve7t33cXUpmBi6ltDwiHgJWAC8Cu+ZTbwJ+WWh7mrpAt8ss6t4X17yxVpIOIs+NHs+JP5tf7DIkAAYPHszYsWM55ZRTAPj0pz/NSSed1OTtx48fz0c/+lF69uzJokWL9ll/4YUXcthhhwHQtWtXfvKTn/DFL36RtWvXklJi+PDhDBo0iD59+jBlyhQqKyu5/vrrD/g8Bg0axEknncTAgQM59thjOfXUUxut8YEHHmDMmDG89tprAEyePJnjjjvugI9ZXzR1KrMURcQngHNTShc3pX9VVVWqrq5u4aokSSq+NWvWMGDAgGKXoSZqaLwi4qmUUlVD/UtmBu5ARcTXgI8CZxW7FkmSpNaUbYBLKf3vYtcgSZJUDKV0E4MkSZKawAAnSZKUGQOcJElSZgxwkiRJmcn2JgZJktR06cU7m3V/ccw1++1z9913M2PGDFJKXHbZZVx11VUArFy5kssvv5xt27bRt29fZs2aRefOnXnyySeZMGEC7du3Z/bs2fTr14+//vWvjB49mh/96EdExD7H2LFjBzfddBNz5syhffv2dOzYkUmTJvHRj36Uvn37Ul1dTdeuXZv13EuBM3CSJKnZrVq1ihkzZrBs2TJWrlzJ448/ztq1a4G6D++dMmUKTz/9NOedd97u7yydOnUqc+bM4ctf/jL33nsvALfffjs33HBDg+EN4KabbmLTpk2sWrWKVatW8R//8R9/1zcs5MIAJ0mSmt2aNWsYMmQIHTt2pG3btgwbNox58+YB8Oyzz3LaaacBMGLECObMmQNAeXk5r776KrW1tZSXl7Nu3Tr+8Ic/MGzYsAaPUVtby4wZM/ja1762+/tKe/TowQUXXNAKZ1hcBjhJktTsKioqWLx4MTU1NdTW1rJgwQLWr1+/e91jjz0GwCOPPLK7/frrr2f8+PHcddddXHnllUycOJHbb7+90WM899xz9OnTh86dO7f8CZUYA5wkSWp2AwYM4Nprr2XEiBGceeaZDBo0iLZt6956f//99zNt2jROPvlkXnnlFdq1awdAZWUlS5cuZdGiRTz//PP06tWLlBKjR4/moosu4uWXXy7mKZUUA5wkSWoR48aNY/ny5SxevJgjjzyS/v37A3DCCSewcOFCnnrqKcaMGcN73vOePbZLKTF58mRuuukmJk2axKRJk7jooou455579ujXr18/XnrppUPiPW97M8BJkqQWsXnzZgBeeukl5s6dy5gxY/Zo37lzJ5MnT+byyy/fY7uZM2dy9tln06VLF2praykrK6OsrIza2to9+nXs2JFx48bx2c9+ltdffx2ATZs28Z3vfKelT63o/BgRSZLUIkaOHElNTQ3l5eVMmzaNLl26ADB79mymTZsGwPnnn8+nPvWp3dvU1tYyc+ZMFi5cCMDVV1/NyJEjadeuHbNnz97nGJMnT+bGG2/kxBNPpEOHDhx++OHcdtttrXB2xRUppWLX0GqqqqpSdXV1scuQJKnFrVmzhgEDBux+XozPgVPT7T1eABHxVEqpqqH+XkKVJEnKjJdQJUk6BDhjdnBxBk6SJCkzBjhJkqTMGOAkSZIyY4CTJEnKjDcxSJJ0CHjzq/+zWffX5urvNqnf3XffzYwZM0gpcdlll3HVVVdx6623MmPGDLp16wbAl7/8Zc466yyefPJJJkyYQPv27Zk9ezb9+vXjr3/9K6NHj+ZHP/oREdGs55AzA5wkSWoRq1atYsaMGSxbtox27dpx5plncvbZZwPw+c9/nmuu2fPO2KlTpzJnzhxeeOEF7r33XqZOncrtt9/ODTfcYHjbiwFOkiS1iDVr1jBkyBA6duwIwLBhw5g3b16j/cvLy3n11Vepra2lvLycdevW8Yc//IFhw4a1VsnZ8D1wkiSpRVRUVLB48WJqamqora1lwYIFrF+/HoCvf/3rvO997+PSSy/lL3/5CwDXX38948eP56677uLKK69k4sSJ3H777cU8hZJlgJMkSS1iwIABXHvttYwYMYIzzzyTQYMG0bZtWyZMmMC6detYsWIFPXv25Atf+AIAlZWVLF26lEWLFvH888/Tq1cvUkqMHj2aiy66iJdffrnIZ1Q6DHCSJKnFjBs3juXLl7N48WKOPPJI+vfvT48ePWjTpg1lZWVcdtllLFu2bI9tUkpMnjyZm266iUmTJjFp0iQuuugi7rnnniKdRekxwEmSpBazefNmAF566SXmzp3LmDFj2LRp0+718+bNo6KiYo9tZs6cydlnn02XLl2ora2lrKyMsrIyamtrW7X2UuZNDJIkqcWMHDmSmpoaysvLmTZtGl26dOHiiy9mxYoVRAR9+/blm9/85u7+tbW1zJw5k4ULFwJw9dVXM3LkSNq1a8fs2bOLdRolxwAnSZJazJIlS/Zpe/DBBxvt37FjRxYtWrT7+dChQ3n66adbpLaceQlVkiQpM87ASZJ0CGjqNycoD87ASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZnxJgZJkg4Bq4d9vFn3d+LP5jep3913382MGTNIKXHZZZdx1VVX8ec//5nRo0fzwgsv0LdvXx5++GG6dOnCnDlzuPnmmznyyCOZP38+Rx11FOvWrWPixIl873vfa9b6c+cMnCRJahGrVq1ixowZLFu2jJUrV/L444+zdu1apkyZwvDhw1m7di3Dhw9nypQpAEydOpWlS5fyyU9+ku9+t+6u2RtvvNEvtG+AAU6SJLWINWvWMGTIEDp27Ejbtm0ZNmwY8+bN49FHH+WSSy4B4JJLLmH+/LrZvLKyMl577TVqa2spLy9nyZIl9OzZk/79+xfzNEqSl1AlSVKLqKioYOLEidTU1HDYYYexYMECqqqqePnll+nZsycAPXv23P19qbfccgsf+chH6NWrF9/5zne44IILvHTaCAOcJElqEQMGDODaa69lxIgRHHHEEQwaNIi2bRuPHiNGjGDEiBFA3Rfan3XWWTz77LPceeeddOnShbvvvpuOHTu2VvklzUuokiSpxYwbN47ly5ezePFijjzySPr370+PHj3YtGkTAJs2baJ79+57bLPrC+0/85nPcP3113P//fdz8sknM2vWrGKcQkkywEmSpBaz6/LoSy+9xNy5cxkzZgznnHMOM2fOBOpm2s4999w9trnjjjv43Oc+R3l5Oa+++ioRQVlZGbW1ta1ef6nyEqokSWoxI0eOpKamhvLycqZNm0aXLl247rrruOCCC/jWt75Fnz59eOSRR3b337hxI9XV1dx6660AfOELX2DIkCG8853v3H2zgwxwkiSpBS1ZsmSftqOOOoonnniiwf69evXi8ccf3/181KhRjBo1qsXqy5WXUCVJkjLjDJwkSYeApn5zgvLgDJwkSVJmDHCSJB2kUkrFLkFN8HbGyQAnSdJBqEOHDtTU1BjiSlxKiZqaGjp06HBA2/keOEmSDkK9e/dmw4YNbNmypdilaD86dOhA7969D2gbA1wLSC/eWewSSsLOOcuLevxnH/UDHyUdfI4/t2lfJVUG9GmB45eNHNwCe81PHHNNUY9f8pdQI6J3RDwaEWsjYl1E3B0R7SLilIhYUXisjIjzil2rJElSayjpABcRAcwF5qeU+gPHAUcA/wqsAqpSSpXAmcA3I8IZRUmtpvPwofR7aDoDFs2l30PT6Tx8aLFLknSIKOkAB/wPYHtK6dsAKaU3gc8Dlxaev1Ho1wHwXZqSWk3n4UPp9aUraHd0d6KsjHZHd6fXl64wxElqFaUe4AYCT9VvSCltBV4C+kXE+yPiGeBp4PJ6gU6SWlT38RdTttddY2UdOtB9/MVFqkjSoaTUA1zQ8MxaACml9MuU0kDgH4HrI2Kfe3AjYnxEVEdEtXfiSGou5d27HlC7JDWnUg9wzwBV9RsiojPwD8C6XW0ppTXA34CKvXeQUpqeUqpKKVV169athcuVdKjYsflPB9QuSc2p1APcE0DHiPgkQES0AaYCDwA9dt20EBHHAMcDLxSnTEmHms3TH2Tn9u17tO3cvp3N0x8sUkWSDiUlHeBS3cdHnweMioi1wO+A7cANwIeAlRGxApgHfCal5J++klrF1ieWsPGOabz+x82knTt5/Y+b2XjHNLY+saTYpUk6BJT8x26klNYDH2tg1YOFhyQVxdYnlhjYJBVFyQe4HBX705lLRZuri3v8E4t8fEmSWkpJX0KVJEnSvgxwkiRJmTHASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZkxwEmSJGXGACdJkpQZA5wkSVJmDHCSJEmZMcBJkiRlxgAnSZKUGQOcJElSZgxwkiRJmTHASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZkxwEmSJGXGACdJkpQZA5wkSVJmDHCSJEmZMcBJkiRlxgAnSZKUGQOcJElSZgxwkiRJmTHASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZkxwEmSJGXGACdJkpQZA5wkSVJmDHCSJEmZMcBJkiRlxgAnSZKUGQOcJElSZgxwkiRJmTHASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZkxwEmSJGXGACdJkpQZA5wkSVJmDHCSJEmZMcBJkiRlxgAnSZKUGQOcJElSZgxwkiRJmTHASZIkZcYAJ0kHudXDPl7sEiQ1s6IHuIhIETG13vNrIuLWes/HR8RvC49lEfGheutmRcSzEbEqIu6PiPJWLl+SJKnVFT3AAa8B50dE171XRMQ/A/8L+FBK6QTgcuC7EXF0ocss4ATgvcBhwKdbp2RJyseARXOLXYKkZlYKAe4NYDrw+QbWXQt8MaX0J4CU0nJgJnBF4fmCVAAsA3q3TsmSlI8oK4Uf9ZKaU6n8r54GXBgR79irfSDw1F5t1YX23QqXTi8GftRiFUqSJJWIkghwKaWtwL8Dn21C9wDSXm3fABanlJbs07nuPXTVEVG9ZcuWv79YSZKkIiuJAFdwFzAOOLxe22rg5L36DS60AxARtwDdgKsb2mlKaXpKqSqlVNWtW7fmrViSJKkISibApZT+DDxMXYjb5Q7g3yLiKICIqATGUjfjRkR8GvgIMCaltLNVC5YkSSqStsUuYC9TgSt3PUkpPRYR7wJ+EREJeAW4KKW0qdDlPuBF4L8jAmBuSum2Vq5Zkkpa2rnTGxmkg0zRA1xK6Yh6yy8DHfdafy9wbyPbFr1+SSp1a844nxN/Nr/YZUhqRv5JJkkHOcObdPAxwEmSJGXGACdJkpQZA5wkSVJmDHCSJEmZMcBJkiRlxgAnSZKUGQOcJElSZgxwkiRJmTHASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZkxwEmSJGXGACdJkpQZA5wkSVJmDHCSJEmZMcBJkiRlxgAnSZKUGQOcJElSZgxwkiRJmTHASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZkxwEmSJGXGACdJkpQZA5wkSVJmDHCSJEmZMcBJkiRlxgAnSZKUGQOcJElSZgxwkiRJmTHASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZkxwEmSJGXGACdJkpQZA5wkSVJmDHCSJEmZMcBJkiRlxgAnSZKUGQOcJElSZgxwkiRJmTHASZIkZcYAJ0mSlBkDnCRJUmYMcJIkSZkxwEmSJGXGACdJkpQZA5wkSVJm2ha7AElS60gv3gnAzjnLefbR2iZtc/y5HSkbOXj38zjmmhapTdKBKekZuIg4OiK+FxHrImJ1RCyIiOMiYmBE/FdE/C4i1kbETRERxa5XkiSpNZTsDFwhkM0DZqaU/qXQVgn0AB4AJqSUFkZER2AO8BlgWpHKlaRsdR4+lO7jL6a8e1d2bP4Tm6c/yNYnlhS7LElvoZRn4M4AdqSU7tvVkFJaARwHPJlSWlhoqwWuBK4rSpWSlLHOw4fS60tX0O7o7kRZGe2O7k6vL11B5+FDi12apLdQygGuAniqgfaBe7enlNYBR0RE59YoTJIOFt3HX0xZhw57tJV16ED38RcXqSJJTVHKAa4xAaRG1u3THhHjI6I6Iqq3bNnSspVJUmbKu3c9oHZJpaGUA9wzwMmNtFfVb4iIY4FtKaVX9u6cUpqeUqpKKVV169atZSqVpEzt2PynA2qXVBpKOcD9F9A+Ii7b1RAR/wisBT4UEf9UaDsMuAe4oyhVSlLGNk9/kJ3bt+/RtnP7djZPf7BIFUlqipINcCmlBJwHjCh8jMgzwK3ARuBc4MaIeBZ4GvgV8PVi1SpJudr6xBI23jGN1/+4mbRzJ6//cTMb75jmXahSiSvZjxEBSCltBC5oZPXprViKJB20tj6xxMAmZaakA5wkqfns+haFNlfDiVcXuRhJf5eSvYQqSZKkhhngJEmSMmOAkyRJyowBTpIkKTMGOEmSpMwY4CRJkjJjgJMkScqMAU6SJCkzBjhJkqTMGOAkSZIyY4CTJEnKjAFOkiQpMwY4SZKkzBjgJEmSMmOAkyRJyowBTpIkKTMGOEmSpMwY4CRJkjJjgJMkScqMAU6SJCkzBjhJkqTMGOAkSZIyY4CTJEnKjAFOkiQpMwY4SZKkzBjgJEmSMmOAkyRJyowBTpIkKTMGOEmSpMwY4CRJkjJjgJMkScqMAU6SJCkzBjhJkqTMGOAkSZIyY4CTJEnKjAFOkiQpMwY4SZKkzBjgJEmSMmOAkyRJyowBTpIkKTMGOEmSpMwY4CRJkjJjgJMkScqMAU6SJCkzBjhJkqTMGOAkSZIyY4CTJEnKjAFOkiQpMwY4SZKkzBjgJEmSMtO22AVIaj3pxTubbV875yzfb59nH61ttuNp/44/t+N++5SNHNxsx4tjrmm2fUk6MM7ASZIkZWa/AS4itjWhz1URsf8//VpIsY8v6eDQefhQ+j00nQGL5tLvoel0Hj602CVJUoOaawbuKuCAAlREtGmmY7+t40tSfZ2HD6XXl66g3dHdibIy2h3dnV5fusIQJ6kkNTnARcTpEfHTiPh+RPw2ImZFnc8CvYBFEbGo0PfDEfHfEbE8Ih6JiCMK7S9ExM0R8XNgVET0i4ifRMTKQt/3FPp9MSJ+FRG/iYhJhba+hePOLLR/PyI6NnR8STpQ3cdfTFmHDnu0lXXoQPfxFxepIklq3IHOwJ1E3WzXicCxwKkppXuAjcAZKaUzIqIrcCPwTymlwUA1cHW9fWxPKX0opfQ9YBYwLaU0CPggsCkiPgz0B04BKoGTI+K0wrbHA9NTSu8DtgKf2fv4exccEeMjojoiqrds2XKApyvpUFHevesBtUtSMR1ogFuWUtqQUtoJrAD6NtBnCHUB78mIWAFcAhxTb/1DABHRCXhXSmkeQEppe0qpFvhw4fFrYDlwAnWBDmB9SunJwvJ3gA/tr+CU0vSUUlVKqapbt24HdLKSDh07Nv/pgNolqZgO9GNEXqu3/GYj2wfw45TSmEb28bd6/RoSwP9JKX1zj8aIvkDaq+/ezyXpbdk8/UF6femKPS6j7ty+nc3THyxiVZLUsOa6ieEVoFNheSlwakT0Ayi8T+24vTdIKW0FNkTExwv92hfuJP1P4NJ675t7V0R0L2zWJyI+UFgeA/y8geNL0gHb+sQSNt4xjdf/uJm0cyev/3EzG++YxtYnlhS7NEnaR3N9kO904IcRsanwPrixwOyIaF9YfyPwuwa2uxj4ZkTcBuwARqWUFkbEAOC/IwJgG3ARdTN+a4BLIuKbwFrg3oaO30znJOkQs/WJJQY2SVmIlPK4Clm4hPp4Sqni7e6jqqoqVVdXN1tNkiRJLSUinkopVTW0zm9ikCRJykw234WaUnoBeNuzb5IkSQcLZ+AkSZIyY4CTJEnKjAFOkiQpMwY4SZKkzBjgJEmSMpPN58A1h4jYArz4d+6mK+CXI5YWx6S0OB6lxzEpPY5J6SnFMTkmpdTgF7kfUgGuOUREdWMfqqficExKi+NRehyT0uOYlJ7cxsRLqJIkSZkxwEmSJGXGAHfgphe7AO3DMSktjkfpcUxKj2NSerIaE98DJ0mSlBln4CRJkjJjgGtERNweEb+JiBURsTAiehXaIyLuiYjnCusH19vmkohYW3hcUrzqD04R8ZWI+G3hdZ8XEe+st+76wpg8GxEfqdd+ZqHtuYi4rjiVH7wiYlREPBMROyOiaq91jkkJ8PUujoi4PyI2R8Sqem1HRsSPC78jfhwRXQrtjf5eUfOIiH+IiEURsabwM+tzhfZ8xySl5KOBB9C53vJngfsKy2cBPwQCGAL8stB+JPB84d8uheUuxT6Pg+kBfBhoW1j+N+DfCssnAiuB9sC7gXVAm8JjHXAs0K7Q58Rin8fB9AAGAMcDPwWq6rU7JiXw8PUu6mt/GjAYWFWv7Q7gusLydfV+hjX4e8VHs45HT2BwYbkT8LvCz6lsx8QZuEaklLbWe3o4sOvNgucC/57qLAXeGRE9gY8AP04p/Tml9Bfgx8CZrVr0QS6ltDCl9Ebh6VKgd2H5XOB7KaXXUkq/B54DTik8nkspPZ9Seh34XqGvmklKaU1K6dkGVjkmpcHXu0hSSouBP+/VfC4ws7A8E/h4vfaGfq+omaSUNqWUlheWXwHWAO8i4zExwL2FiPjXiFgPXAjcXGh+F7C+XrcNhbbG2tUyLqXuryNwTEqRY1IafL1LS4+U0iaoCxRA90K749SKIqIvcBLwSzIek7bFLqCYIuInwNENrJqYUno0pTQRmBgR1wNXArdQN526t/QW7ToA+xuTQp+JwBvArF2bNdA/0fAfKI7JAWrKmDS0WQNtjknr8+dSHhynVhIRRwBzgKtSSlsjGnrp67o20FZSY3JIB7iU0j81set3gR9QF+A2AP9Qb11vYGOh/fS92n/6dxd5iNnfmBRuDvlnYHgqvFGBxseEt2hXEx3A/5P6HJPS8FbjoNb3ckT0TCltKqO6spkAAAFxSURBVFyO21xod5xaQUSUUxfeZqWU5haasx0TL6E2IiL613t6DvDbwvJjwCcLd6gMAf5fYdr1P4EPR0SXwl0sHy60qZlExJnAtcA5KaXaeqseA/4lItpHxLuB/sAy4FdA/4h4d0S0A/6l0FctzzEpDb7epeUxYNcnFFwCPFqvvaHfK2omUTfV9i1gTUrpq/VWZTsmh/QM3H5MiYjjgZ3Ai8DlhfYF1N2d8hxQC3wKIKX054i4nbofmAC3pZT2fgOr/j5fp+6uxh8Xpr2XppQuTyk9ExEPA6upu7R6RUrpTYCIuJK6IN0GuD+l9ExxSj84RcR5wNeAbsAPImJFSukjjklpSCm94etdHBExm7qrMl0jYgN1V3CmAA9HxDjgJWBUoXuDv1fUrE4FLgaejogVhbYbyHhM/CYGSZKkzHgJVZIkKTMGOEmSpMwY4CRJkjJjgJMkScqMAU6SJCkzBjhJkqTMGOAkSZIyY4CTJEnKzP8H87ab8cDPt0AAAAAASUVORK5CYII=\n",
"text/plain": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# Add interval percent widths\n",
"alphas = [ 0.01, 0.05, 0.1] \n",
"widths = [ '99% CI', '95%', '90%']\n",
"colors = ['#fee08b','#fc8d59','#d53e4f']\n",
"\n",
"for alpha, color, width in zip(alphas, colors, widths):\n",
" # Grab confidence interval\n",
" conf_ints = res.conf_int(alpha)\n",
" \n",
" # Pass current interval color and legend label to plot\n",
" plt.hlines(y = conf_ints.index, xmin = conf_ints[0], xmax = conf_ints[1],\n",
" colors = color, label = width, linewidth = 10) \n",
"\n",
"# Draw point estimates\n",
"plt.plot(res.params, res.params.index, 'wo', label = 'Point Estimate')\n",
"\n",
"plt.legend(loc = 'upper right')"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 90 and 95% bands\n",
"You are looking at a 40-day rolling average of the $NO_2$ pollution levels for the city of Cincinnati in 2013. To provide as detailed a picture of the uncertainty in the trend you want to look at both the 90 and 99% intervals around this rolling estimate.\n",
"\n",
"To do this, set up your two interval sizes and an orange ordinal color palette. Additionally, to enable precise readings of the bands, make them semi-transparent, so the Seaborn background grids show through."
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"int_widths = ['90%', '99%']\n",
"z_scores = [1.67, 2.58]\n",
"colors = ['#fc8d59', '#fee08b']\n",
"\n",
"for percent, Z, color in zip(int_widths, z_scores, colors):\n",
" \n",
" # Pass lower and upper confidence bounds and lower opacity\n",
" plt.fill_between(\n",
" x = cinci_13_no2.day, alpha = 0.4, color = color,\n",
" y1 = cinci_13_no2['mean'] - Z * cinci_13_no2['std_err'],\n",
" y2 = cinci_13_no2['mean'] + Z * cinci_13_no2['std_err'],\n",
" label = percent);\n",
" \n",
"plt.legend();"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This plot shows us that throughout 2013, the average NO2 values in Cincinnati followed a cyclical pattern with the seasons. However, the uncertainty bands show that for most of the year you can't be sure this pattern is not noise at both a 90 and 99% confidence level."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Using band thickness instead of coloring\n",
"You are a researcher investigating the elevation a rocket reaches before visual is lost and pollutant levels at Vandenberg Air Force Base. You've built a model to predict this relationship, and since you are working independently, you don't have the money to pay for color figures in your journal article. You need to make your model results plot work in black and white. To do this, you will plot the 90, 95, and 99% intervals of the effect of each pollutant as successively smaller bars."
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# Decrase interval thickness as interval widens\n",
"sizes = [ 15, 10, 5]\n",
"int_widths = ['90% CI', '95%', '99%']\n",
"z_scores = [ 1.67, 1.96, 2.58]\n",
"\n",
"for percent, Z, size in zip(int_widths, z_scores, sizes):\n",
" plt.hlines(y = rocket_model.pollutant, \n",
" xmin = rocket_model['est'] - Z * rocket_model['std_err'],\n",
" xmax = rocket_model['est'] + Z * rocket_model['std_err'],\n",
" label = percent, \n",
" # Resize lines and color them gray\n",
" linewidth = size, \n",
" color = 'gray'); \n",
" \n",
"# Add point estimate\n",
"plt.plot('est', 'pollutant', 'wo', data = rocket_model, label = 'Point Estimate');\n",
"plt.legend(loc = 'center left', bbox_to_anchor = (1, 0.5));"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"While less elegant than using color to differentiate interval sizes, this plot still clearly allows the reader to access the effect each pollutant has on rocket visibility. You can see that of all the pollutants, O3 has the largest effect and also the tightest confidence bounds"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Visualizing the bootstrap\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### The bootstrap histogram\n",
"You are considering a vacation to Cincinnati in May, but you have a severe sensitivity to NO2. You pull a few years of pollution data from Cincinnati in May and look at a bootstrap estimate of the average $NO_2$ levels. You only have one estimate to look at the best way to visualize the results of your bootstrap estimates is with a histogram.\n",
"\n",
"While you like the intuition of the bootstrap histogram by itself, your partner who will be going on the vacation with you, likes seeing percent intervals. To accommodate them, you decide to highlight the 95% interval by shading the region."
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {},
"outputs": [],
"source": [
"# Perform bootstrapped mean on a vector\n",
"def bootstrap(data, n_boots):\n",
" return [np.mean(np.random.choice(data,len(data))) for _ in range(n_boots) ]"
]
},
{
"cell_type": "code",
"execution_count": 23,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"pollution = pd.read_csv('./dataset/pollution_wide.csv')\n",
"cinci_may_NO2 = pollution.query(\"city == 'Cincinnati' & month == 5\").NO2\n",
"\n",
"# Generate bootstrap samples\n",
"boot_means = bootstrap(cinci_may_NO2, 1000)\n",
"\n",
"# Get lower and upper 95% interval bounds\n",
"lower, upper = np.percentile(boot_means, [2.5, 97.5])\n",
"\n",
"# Plot shaded area for interval\n",
"plt.axvspan(lower, upper, color = 'gray', alpha = 0.2);\n",
"\n",
"# Draw histogram of bootstrap samples\n",
"sns.distplot(boot_means, bins = 100, kde = False);"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Your bootstrap histogram looks stable and uniform. You're now confident that the average NO2 levels in Cincinnati during your vacation should be in the range of 16 to 23."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Bootstrapped regressions\n",
"While working for the Long Beach parks and recreation department investigating the relationship between $NO_2$ and $SO_2$ you noticed a cluster of potential outliers that you suspect might be throwing off the correlations.\n",
"\n",
"Investigate the uncertainty of your correlations through bootstrap resampling to see how stable your fits are. For convenience, the bootstrap sampling is complete and is provided as `no2_so2_boot` along with `no2_so2` for the non-resampled data."
]
},
{
"cell_type": "code",
"execution_count": 24,
"metadata": {},
"outputs": [],
"source": [
"no2_so2 = pd.read_csv('./dataset/no2_so2.csv', index_col=0)\n",
"no2_so2_boot = pd.read_csv('./dataset/no2_so2_boot.csv', index_col=0)"
]
},
{
"cell_type": "code",
"execution_count": 25,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAWQAAAFgCAYAAACMigM2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nOy9a6ylWXrf9Vvveq/7ei51657uuY9NYjM4divYCaCYsZERHkWBMATJIPiQEQh5EhCghA/j0SBAEReTERBhc4tEMLR8ETiCkWGCHYJkK217aF/GY2dmuqeruqvqXPftva4LH9a7d1X1VE1XTdc+tav6+Unj47PPPvWuU6fr/673Wf/n/yjvPYIgCMKTJ3rSCxAEQRACIsiCIAg7ggiyIAjCjiCCLAiCsCOIIAuCIOwI8ZNewN382I/9mP/iF7/4pJchCILwuFEP86ad2iEfHx8/6SUIgiA8MXZKkAVBEN7LiCALgiDsCCLIgiAIO4IIsiAIwo6wVUFWSv2bSqnfU0r9rlLq55RS+TavJwiC8DSzNUFWSr0P+Azwkvf+ewEN/IVtXU8QBOFpZ9slixgolFIxMADe3PL1BEEQnlq2Jsje+xvAfwJ8E3gLmHnvf+Xt71NKfVop9YpS6pWjo6NtLUcQBGHn2WbJYh/4s8CHgOeBoVLqJ97+Pu/9z3jvX/Lev3T58uVtLUcQBGHn2WbJ4keAb3jvj7z3HfCLwJ/a4vUEQRCearYpyN8EflApNVBKKeATwFe2eD1BEISnmm3WkH8D+Hngt4Df6a/1M9u6niAITwmvvgw//b3wub3w8dWXn/SKdoatpr15738K+KltXkMQhKeIV1+GX/4MdFX4fPZG+Bzg4596cuvaEaRTTxCEi+NLn78jxmu6KrwuiCALgnCBzK4/2uvvMUSQBUG4OKYvPNrr7zFEkAVBuDg+8VlIintfS4rwuiCCLAjCBfLxT8EnvwDTFwEVPn7yC3Kg17NTM/UEQXgP8PFPiQA/ANkhC4Ig7AgiyIIgCDuCCLIgCMKOIIIsCIKwI4ggC4Ig7AgiyIIgCDuCCLIgCMKOIIIsCIKwI4ggC4Ig7AgiyIIgCDuCCLIgCMKOIIIsCIKwI4ggC4Ig7AgiyIIgCDuCCLIgCMKOIIIsCIKwI4ggC4Ig7AgiyIIgCDuCCLIgCMKOIIIsCIKwI2xNkJVS362U+vJd/5srpf7ytq4nCILwtLO1qdPe+68C3weglNLADeCXtnU9QRCEp52LKll8Avia9/71C7qeIAjCU8dFCfJfAH7ufl9QSn1aKfWKUuqVo6OjC1qOIAjC7qG899u9gFIp8CbwPd77W9/uvS+99JJ/5ZVXtroeQRCEJ4B6mDddxA75nwZ+653EWBAE4b3ORQjyv8gDyhWCIAjCHbYqyEqpAfCjwC9u8zqCIAjPAluzvQF470vgcJvXEARBeFaQTj1BEIQdQQRZEARhRxBBFgRB2BFEkAVBEHYEEWRBEIQdQQRZEARhRxBBFgRB2BFEkAVBEHYEEWRBEIQdQQRZEARhRxBBFgRB2BFEkAVBEHYEEeT3Gq++DD/9vfC5vfDx1Zef9IoEQejZatqbsGO8+jL88megq8LnszfC5wAf/9STW5cgCIDskN9bfOnzd8R4TVeF1wVBeOKIIL+XmF1/tNcFQbhQRJDfS0xfeLTXBUG4UESQ30t84rOQFPe+lhThdUEQnjgiyO8lPv4p+OQXYPoioMLHT35BDvQEYUcQl8V7jY9/SgRYEHYU2SELgiDsCCLIgiAIO4IIsiAIwo4ggiwIgrAjiCALgiDsCCLIgiAIO4IIsiAIwo6wVUFWSu0ppX5eKfUHSqmvKKV+aJvXEwRBeJrZdmPIXwe+6L3/80qpFBhs+XqCIAhPLVsTZKXUBPgngH8FwHvfAu22ricIgvC0s82SxYeBI+C/V0r9tlLqv1FKDd/+JqXUp5VSryilXjk6OtricgRBEHabbQpyDHw/8De8938CWAF/5e1v8t7/jPf+Je/9S5cvX97icgRBEHabbQrydeC69/43+s9/niDQgiAIwn3YmiB7728Cbyilvrt/6RPA72/reoIgCE8723ZZ/CTwt3qHxdeBf3XL1xMEQXhq2aoge++/DLy0zWsIgiA8K0inniAIwo4ggiwIgrAjiCALgiDsCCLIgiAIO4IIsiAIwo4ggiwIgrAjiCALgiDsCCLIgiAIO4IIsiAIwo4ggiw8mFdfhp/+XvjcXvj46stPekWC8Eyz7SwL4Unz6svwpc/D7DpMX4BPfBY+/qmH+75f/gx0Vfh89kb4HB7u+wVBeGRkh/wssxbV2RuAvyOqD7PT/dLn74jxmq4KrwuCsBVEkJ9l3o2ozq4/2uuCILxrRJCfZd6NqE5feLTXBUF414ggP8u8G1H9xGchKe59LSnC64IgbAUR5GeZdyOqH/8UfPILMH0RUOHjJ78gB3qCsEXEZfEssxbP78Rlsf5+EWBBuDBEkJ91RFQF4alBShaCIAg7ggiyIAjCjiCCLAiCsCOIIAuCIOwIIsiCIAg7ggiyIAjCjiCCLAiCsCOIIAuCIOwIW20MUUq9BiwACxjv/UvbvJ4gCMLTzEV06v2w9/74Aq4jCILwrmmN5WzZcHlaECl1odeW1mlBEATAWMfpsmFWtgDEOuJwnF/oGrZdQ/bAryilflMp9en7vUEp9Wml1CtKqVeOjo62vBxBEIR7sc5zsqh57fZiI8aJjsgSTWss3vsLW4va5sWUUs97799USl0B/k/gJ733f/dB73/ppZf8K6+8srX1CIIgrPHeMytbThYNrtfBSCkOxxmjPOZ02TIrWy6Nc/ZH2bu93EPVPra6Q/bev9l/vA38EvAnt3k9QRCEh2FRdbx+tORoXuO8RwEHo4z3Xx7hvOf1o+Vmt7yo2gtb19ZqyEqpIRB57xf9//9PATIhUxCEJ0bZGI7nFY1xm9emg5T9YUrVWt44XmLdercMB6Oc6TC9sPVt81DvKvBLKpxSxsD/5L3/4havJwiCcF+aznI8rylbg/OeSCmGWcylSU5nHG+elbR3ifTeMOVglKOjZ8Rl4b3/OvCPbOvPFwRBeCc66zhZ1CyqDuc886qlM47v/cABaay5PauoWrt5/7hIOBznJPrJ9MyJ7U0QhGcO64KF7XzV4r1nWRtWTcekSBnlCW8cLUkTvXl/kWouTQryu157EoggC4LwzOC853zVcLZscD7UjBdVyzBPuDTOWNaG1lgmRagLp3HEpUnOMEue8MoDIsiCIDz1eO9ZVB0nixrjPE1nmZUNWaI5nOSUtWFRdUwGKTqK0JHi0jhnXCSoC+7G+3aIIAuC8FSzqjuOFzWtcXTGcV426EhxOM6pWst81TIqEtI4JVKwP8rYG2YX3hb9MIggC4LwVFK3huNFTdVarHPMyhbvYG+QYpxntmoZZDGjvv15Okg5HGfoaHdDLkWQBUF4qmiN5WTRsKw7XF+qaDvLZJiiCE0feRpz0AvxKE84HGek8ZM9sHsYRJAFQXgqsM5xsgjhP2vnRN12jIqUYRazqDpirdgfZSilyBPN5UlOnj49Mvf0rFQQhPckznvOlsE54QnOiWXVMsgTDsYFy36HPB2kRJEi0cE5Mcp3wznxKIggC4Kwk6zDf06XDbZ3TszLlizRXJrkrBrDvGwZFwmxjjYHeZMdc048CiLIzxKvvgxf+jzMrsP0BfjEZ+Hjn3rSqxKER2ZtYetscE7MyxYdKQ7GOXVrmZUdwzxmXGgUwTmxP9pN58SjIIL8rPDqy/DLn4GuCp/P3gifg4iy8NRQtYajeU3TWazzzMoGPEyHKcY6ZmXDII03cZjTQcrBKCN+Qq3OjxsR5GeFL33+jhiv6arwugiysOM0neV4UVM2ZuOc6IxlMgjOiXnVksWag1FwTqyDgZ4G58SjIIL8rDC7/mivC8IOcHf4j/eeVWOoWsO4SBhmMcuqI4oU+8PgnMh650TxFDknHoVn86d6LzJ9IZQp7ve6IOwY1gXnxPnqjnNiVXcM84SDUc6ybml82CGvnROHfavzs8yzUXgRwgFeUtz7WlKE1wVhR/C9he212wvOVg11ZzmaV1jnOBxnWBcO8IZZwnQYasOXJzkfuDx65sUYZIf87LCuE4vLQthR5lXLyTyE/3TWMV+1aB2sanVrOS87Rm9zTuwNswsPiX+SiCA/S3z8UyLAws6xajpO5jWNcVjnmfez6vZGKZ11nK8aijTmoHdOTPqQ+GfFOfEoiCALgrAV6s5yctfYpGXV0VnHpC89zMp7nRODLObSOCd7wiHxTxIRZEEQHiutsZwuGhb1HedE3TsnBlnMsu6IlNpEYGZxxKVJwSDbPTmqW0Osowvbre/e34AgCE8ld4f/QGjyWNUdwyzhYJyzqFq8D9M61s6Jg3G2md6xK6xvImfLcOi4P8y4NMkv5NoiyIIgvCvW4T/nqzA2qeksy6ol6+vCZRsyKEZ5yJyIFByMcvaG6U5lTjgf6tvnq5bO3plAXXfmwtYggiwIwneE9555nzlhe+fEomyJdcT+OKe5yzmxTl7bH4bMiV1yTljnOF8FIXbeb14fZDH7w+xCSykiyIIgPDLLuuN4HsJ/Ns4JBdNhhrGW2aoJIfG9c2LcOyeSHXJOtMZytmyYV909r0+KhL1hhnWe02WN9xnDC4ryFEEWBOGhqVrD8bym7uzGOWGcY1KkYdBo2RLHEftr50QaczjJyXfIOVE2hrNVQ9ncVYrwoCPQUcQwi7k1q2g6C4DzjQiyIAi7Q2ssx/OaVWM2h15NZxkXCToK0zoipZgMUyKlSOMQEj/MdqO7znvPou44XzY05k59WAFKgfWeeWWZVw3nZbspU+SJ5mCcXdg6RZAFQXggpg//WT/WV62hbAzDLGYwylhWLd7DuEjRkSLuQ+LHOxISvy6nnK8ajPP3fC1SYK1j1R86WucZ5jFJHDFIYw7G2YWHGL3j1ZRS/xDwPuA3vPfLu17/Me/9Fx/i+zXwCnDDe//j72axgiBcDNZ5zld3xia1xrKsOrIk1IVXjaFetQyLhKR3TqxbnXchJH7dATgvW+7WYdX/H+tCxGfoGlQhXS5PmBQJB6Psnjl83vsLu7l8W0FWSn0G+DeArwD/rVLqL3nv/9f+y/8h8I6CDPyl/vsn72ahgiBsn/XYpJNFg/Nr50RHEofhoXVnOV+FmuraObE3DCHxOnryB3Z1ZzlfhqaUNd57dKRw3mM9LMuOedWio4jpMGOYxUyKlP1RtukStM6F8VGLhoNxvjmc3DbvtEP+i8APeO+XSqkPAj+vlPqg9/6v099svh1KqReAfwb4D4B/612uVRCELTKvggCtnROLqkWpflpHbw3LU705sBvnCQfjjPT3f+GJh1qtmo7zZUvZ3usZXm9sOxt+nmXVkSaaw3HIVJ4UCfujjDTWrOqOm+clWaw5mtcs646yDfnMuyLIel2m8N6/ppT6MwRR/gAPIcjAfw78u8D4Xa1SEIStUTaG43lFY9zGOeG83+yAF1XwFq8bOYpUc2lSBOfEExwd5vvJImerhvaugzo8RFE4qLPWsyg7lk0XAozGGcZ6rk0L9sc5caQ4L1u++uY5x/MQlJ8nGt2XYbI4Ir1Aq947CfJNpdT3ee+/DNDvlH8c+O+Af/jbfWP/vtve+9/shfxB7/s08GmA97///Y+ydkEQ3gV3h/+snROtsYzzhCiKWNYdCkJI/No5Mc7vtYA9gdFh63LC+SocxK1RCpQHp6A1jkXVsmoMozzhcJRRd47WOMZFQppqThY1109WHM+qUBM3ljSOGGQxWRKmWOdJzN4FuiyU9/7BXwwlB+O9v3mfr/1p7/3/+22+9z8C/iXAADmhhvyL3vufeND3vPTSS/6VV155hOULgvCodNZxMq83ddaqNVSNYZjHpLFmWXVYT29pU+jeOTG5n3Pic3vA/TREwefOH+u6W2M5X7XMy3ZzRe89Wim8Au+hM5ZF3VG3wZIXxxFVbYgixbj3Sh/Pa8q2o+0cq8ZgvadIYg7GKamOSGLNIIs38/oGacz7DofvdvkPdSr4bXfI3vvrAEqpHwa+h/A3//ve+//724lx/71/Ffir/ff/GeDf/nZiLAjCdrHOcbpsOF+F8J/WWJa1oUg0B+OcZd1RNfc6J/b6VucHOicuYHRY3RrOVi3Luw7q8B4VKUDhgK6zzKoOY0OTSpFqVrXF+/AzNMbw5umSedVhrdt4kcd5wmSQUqS93S2LNz/ruEjYG6T3OC62zTu5LN4H/CJQA79JUPlPKaX+GvDnvPc3tr9EQRDeDc7fsbA5H7zFi6ojiSP2hylNZzlb1vc4J6aDlMPxQzgnPvHZe2vI8FhGh709ce2uL6CUwqHAQ2Msi7LD45kUKdZ5Vk2oA++PUhZ1x9dvzVjWBusc1rOZRjIpYvI0IU80aawx1lE2hukgwXtFnmhOlw3Tgd+ZTr3/Avgb3vv/4e4XlVL/MvBfAX/2YS7ivf9V4FcffXmCIHynrA+9ThZhbJJ7u3PCOs7Lthev4JwY5QmH42zzuP6OPObRYQ9KXMN7UGpTqmiNYV516EgxGSS0JtSVB1nMdJByvmp4/WhJ3Ro84dE+0YqDYca0SMkSTdHvhqvGcLSqqDuDc57OWprWcbKo2B/lOH9xgvxONeSveu+/+1G/9p0iNWRBeDys6o7jRU37NufEuEjwPoQDxTrkNigVdoOXJ/mFPp7fzYMS195O3RqWdbep89a9LW2Qhp/jeFFzsgihR+t69yDVTIuUvVFGGkdkSRxsfWXLsumw1ofJ161hWbUcjHMORxmjPGFUpOwPs8cxYPXd15CB+94mlVLRg74mCMKTo24Nx4uaqg2P+cu6wxjHqD+QW/V12LVzItEhc2J0QTvAt3O/xDXnPbo/pFsbiavGsGoNeaKZDlNWteF00TDIQrnhxsmKeR3qw5GOiKOIYZ6wN0wZ56E0oSNF3RlunZeUrQmlD+dZ1h2rpiPREdNBxgcvjdgb5ewPL7Z+DO8syH9bKfWzwF/23q8AlFJD4KeB/33bixOErfLqy0+8oeFx0RrLyaLZHHxVraFpLUUeM+zHJll3xzkRKcXhOGM6eDIh8fdLXPO9f1h58Cg8nrI2tMaQpzF7g5RF1VLWof4Nnq/dmlPWBgXoOKLIEgZpCAQa5ylpovF9DX1Zd7TGkcQaYz3LqqG1jiLRXJ0UpIlmmCdc2RswHaQs647b85q9QcpkcDFTTd5JkP8dQov060qp1wmlmA8AfxP497a8NkHYHk+woeFxYmxwTqzHJrXGUtaGLNXsjTJWtQnOiTw4J9YHWnvDiw+Jf1DiGoTnea883oc1rergmCiymDROmFcdnXForVi1hpPjmqqzJJEiTTR5ohllofFjmCfoKKI1lptnK5aN2dyEOus5X5ZoHcT7INUkOmJSJFzdK0i05tas4v977Zg0jjhbtrzvcMT3f/jShfwdvZMgfx/wnwGfBT4K/DDw40AKjIDTra5OELbFE2hoeJy8PfzHWLepC++NMuouOBQGWcwwD40Nkz4k/qIGdt691kXVcra8N3HNAxGhRIFSeA9lE2rdRRoTa8Vs1dKaUH6ZVx3nq9DancaacW/POximjIuMUV/nPV/VzFYdjbFkiUZHEau6ozOWNA4WP60jskSzPwxukqazXD8pOV1WnC7DWsGTpzFFdnHV2XcS5P8a+BHvfaWU2gf+CvCTBKH+GeDPb3l9grAdZtcf7fUd4e3hP+saqFLBqmas43zZkKWa/T5/YZjFXJrkD++ceEwYGw7qZmWzSVzz3m8S1xRq44CoGoMnCHFjLEfzitY4rHUs6o5FFSZYZ2nMqEjIY810kDEdJmRJTGcMb52uWDYdkYpI4ggdRSz6JpI81UwGKYo7o5mKVDOvWv7wzRkni5qzVcuibOmMAxUOQEPn3sXV1x8my2K9C/4XgJ/x3v8C8AtKqS9vd2mCsEUuoKHhcbO2sHXW4XvnhAeGebKZb6cjxbTPnMh658RFZ/o2neVs1bC466DOu3UjB5uDOus8rbEoIE8iqtZx42RJayyd9SwbQ1m3REozyGOKNA6liSJhf5gR64hZ2fLW6ZyqsxRp6K5rOktddcRReG8UhSaXYRYzHWQoPOdly+tHNSeLkJVctaFjzzlPpBRZrBnlaX/guSONIYBWSsXeewN8gj5z4iG/VxB2ly01NGyDsgnOifVIoXV9dZgnRJFi2QvfuEg2zol1SPxFcr/EtZAlzMb0pZQKDRp9h0aeaBZ1x43TkqazGGv7TA1HrBX744JcR2SZZpyHwzXnHEezinnVEUUhY6PIYpquQxGR6Gjjjkh06L4bZpq6cxzNK04WNaerlvNljbHBFui9J4pCkFCearIk7htl0gudevJOovpzwK8ppY6BCvh/AJRSHwVmW16bIGyPx9zQsA2aznK8qDdOhLoNY5MGWcygd044H5o5npRz4kGJa/eGuiuUAndXp5zWEeerhtvnJbVxoX7cGpx1ZGnElWlIk0t0xGQQwuOXdcdrR3OaxpJnMUUWY6zDWIdSUKR3nBBZL6xpoqlbw/XThtN5vbEEGmuwzuMcaB12xMMiYTpIuTTOGRcp00HKMIsvrCkE3qExBEAp9YPAc8Cv3GV9+y5g5L3/rce5GGkMEYQ+/GdRbx75W2OpWkuWRORJTNmnst3tnFhnTlyUc8I6z6xsviVxDXwoCt91Q3DO9V9ReO+4Pau5dV5inMM7aPtd6jhPGBcxaRxyJSZFRqzhbNlwtmo30Z9KKYy1JFGEUnfq0HhPrFU/D09hrOd4XnF7UTNbNjRdKEu0xqE8ZGnMuIiZFBnTYRBfD1ydFoDiw1fHjIvHZnd7LI0heO9//T6v/eF3siJBEB6MdZ6zZcP56o5zYtWYvmEhpekM58uGIovZG4YDu3HvnEguyDnxoMS1YFtTqHBat/7KRpzbzvLWWcntWdk3fEBnQjkjNG+kRJEiiUMqm3OOW7MVZRM8yOs6eaQUUQRJpOlc+ONdn/hWZHFfwjHcPC+5eV6yqg3GOawF6x1pHKaETIuU6bCfEqIjOuNxhD9foUjiiLfOSt44XjIZpLxwOLqQv1+pAwvCE8Z5z2zVcrqscR5cH5CjVLCqGeuYrVqSOFjaIERCXprkm5FD2+Z+iWve9/WH3jax1uEoCsE/6wGjN05XnCwbFBBrRWccaaK5upcxzBKcD7XkItUsm47Xby/whJ3uZJChgDSJUAqMcRgHnfNY50iiqD90U9w8XfLGaclZWWOsx3tw1hNF9F17A/aGKZfGBfvDBOegMY7Wut4I7Wlaxx/dPOdkXgOKxlg+em0igiwIzzpvD//x3vc+3HudE5EKATpKKbI44tKk2Iyp3/b67pe4FnbEQX7VXU/iiVY4v+4arLl+vNrMrou1wjhHojXP7Q/RUYT3wU8cacVs2XJ7VhLHoVtuHYifxhHWOerWbcoN3kORRhRpwvmy5g/fmnE0r/r8Co91YKwP9ra9jCvjgsNxxuVJmBAyrw3zKoQOWeuo2o6yNpyXLcvabH7WLA6t1I2xXBQiyILwBFg1HcfzenMQtqo7rPebsJ/1tI61cyKOFIeTnMnjq2k+kAclrm12xLDxEUcK4jjCu3DzuH1ecuOspGwMWR9rqVTIHd4fpUAQ7ViDc4rb8wo85FnMeJCRaEXRj1CqW8uiNnTW0XQG74IQt9bztVsLjmbhgC6KABd25ErBlWnBtemA/XHoSBy+8Ws0v/I/MltVmMEV2j/256gufx/zsmVRGsrO3HMgmaeaw1HGR65N+PC1Kdf2iq3/na8RQRaEC6TuLMfzahP+U7eGzjjyVBPriLIJNc9R3/4bKTgY5UyH6YND4h8T90tcu/vQf+OaUBD3drO6Cxa0G6ehPmyMJ091WL9W7BcpwyLBOY/ta82NsZyvTEibyxO0UuSpZpzHIWOiMdRlCPwpa4OKFLEOwUh/dLOmrMPhnFJhbc55poOM9x0MuDQpyBLNJE9QkWL1B7/Gzb//t+is5YSr3C4PmH35d4ifi/DTF1nfYUIEacZHrk742HNTruwVF95IAyLIgnAhvD38pzWWugutveNBGtLMmpZhljCMg81qb5hyMHqIkPjHsLb7Ja5FrM0Sd24EWRxGHM3KhhunS26crDhdtaGMkMUMsohMR+yNUrJY01lP1Rq8h6rf5eZpzHSYkeqIcRGTJzHLpuP2rGZetcyrjrazxDpCKc9iZZj3trqof00DWRJzbW/A1b2CNIkZZJoijTHGcVa21K3h9Lf/DjftBzlnnw6NQ6OsIrn1R0wufTCI8JUJH3t+ytW9wUaEbV9CCgYSv/XfwRoRZOGZSj3bNaxznCzuhP9YF6ZSxDpiUqQ0xjJbNSHNbO2cyBMOJ9t3Ttw/cc3fcazdtSMfpJqozxt+66zk1qzkfNVtbGZ5ohlmMaN+Z9q0lnk/Usk6TxQpBmmyGSI6HaQ4F+bbfW0+Z1F1mxCgRCu89xzNK5rOAZ44jsiVIooiJoOUK5OcyTAjjhSTdYNM3fHWcsnJouFoXnE0r2mbF/EoNB6FY0TJIWd8rPs63/Wn/zWe3x+QxnH/uwmlmrNVw8mi3szg2xtmXJleTNlCBPm9zjOSerZrOH/Hwna3c0IrxShPgo+3d05MeyEuUs2lSWiI2BYPSly7tx8h1IYVMCoSOmO5frri5lm5aVSJ4xDqk2eaUR6TJ0k/PinUY41zRKg+0lJvBLtIY06XNV+5fsZpf1hobRBsHSma1rC0Hus8cawo0ghURKzgcJJzOM7JkpgsjsgTjfWeo0XN0aLmeFZxNKto+lD+WEfoJGXYnbPPGR/l63yMb/Iib5FNr8HlyUaE18H2VWtDA45xxFFJlsTUnRVBFi6Ipzz1bNdYOyNOFvXmsbdsOkDd5ZxoiVS0cU6kcQiJ32aL7lp4zlf3Jq65u3bE6xpxpBTjImZVG756/Yxb84rzVUPdObJYc2VaMExj8ix00tVtKHkEgfckOmKUJ2RxRJpoijjGeseNsxUn84aqM9St7X3FkGiN85527W5INM6Hm8UwC2WbySA0vRSpJlIh1+Mbt+ec9N13y8bgbdhJh9p02IV/ZPIBPvo7/wsfsF+jIGR/LPQlyj/1Wb0SIqcAACAASURBVE5uhlChsjHUraE2jlXVUbUdxsPHrk3CtO0LykIGEWThKU0920WWdXBOrJ0JZRNmtBVZglLBSeG93zRB6Ej1bbrJ1lqdH5S4tm5tju46qEv6duPTRc1X3zzjdNWyLFta6xhkMS8cFIyKmEhFREqxagynbajtxlHIk4h1RBIpYh2RxYrOeL5xNud82VC2tt95QhZrslhjCULsvEdFEao/rDsc5RyMwkippD9AXDUd/+DmkpNFL8K1wViH1hF5rBkOE6aDhA9eHvHR5/f44JUxw+zj+A9Zlv/Xf8wfLAw3Bt/D+Uf/WTL9/fjr5zSdYdH/jM5BpCGLY3IVUuE+dGV8IRbDNSLI73WewtSzXaNqDcfzeuNfrdtg1QqZvkFIbB8GtHZOrEPit+WcuF/imuttYR6I7jqkypMIHSneOq/45tGCWdlR9W6PSZHygWlOGkeAorOOVduyagy+75DLkhgdhZ11sW7wqAxfP13esc55SGLFpXFGpKA2lrILNeNwY/BkcQjV3xuE2nCiFVVnuH5ac3sRQoEWtaHrHFGkyJOI/dGAvUHCC5eGfPTKlA9fm2zanZd1y1eun3GdH+Lsh/5nqqajMY66NXBrjtYR3oeDyiKNN3nMJ1WFt54//sI+k0HKO8VLPE5EkN/rPEWpZ7tGayzH85pVfyjWGUfdBf/tOE2pW8Oq7hhkCUlfjpgOQiD6tk7t75e4Zl2wmynlUSraeCaKRNM5x9dvLbk9K1nWHVVfSjgYZVybFqBC1sSqCWWGZdNhjSdONGkUxLjIdH+wFnG0qPnGrfkmzlIRDimLRPd+YruJukh0WFvwKOcM8gStoGott84rjhY158uWZdVSdjYECCWa/b2c/UHOcwc5H7oy5SPXxuyPQo13WbX83hunXD9ectrvyltjqZqOsrVYZ4mU5vI0J4sjytaFYKQudOvFWhEROvi+cuOMVWt4bm/Ax57f28rv6+2IIL/XeQpSz3YN04f/rG1i1jmq1hJHIYehvcs5sT6wG+UJh+NsK97WByWurRsllKLfiYdI+CTWlI3h927OOFk2VK2lag1JHHFtb8DlSUbXTyCpjWNeBoH3zlNkMaNBEux5ecIwi2mN5da84mRe01lw3hEpxTDRZHFMay2lsSgfbgrWhjXsDQdMBwlpolnVLW+dLbl9XoeGjboNB37OkyWaq9OCg3HO1WnOBy6N+fDVCZemBZFSLKqW3/vmCa8fLTlehCeV1jhWdUfdGqwPQhvcG6Ed+rWjObHSZIkmjyOyVFO1hrNlh7HhptG9aXnjZMWHrkxEkIUL5OOfEgF+CN4e/uNc8Niq3jlhrGNWNiRab4Q470PitzG9+H6Ja+v6MKh7kt8UYUzSbNly46xkVrW0bfBCF2nMx56bMEhDxOUbJ0vKxrJqOprWhUaLQQh3L7KYSRHCgBZ1wzduz5n1ZQkPGxta6MjztM7iCDcxPAyzhP39MK2j6Sw3z0tOF0GEl21H2YQ4zTSO2B9lXJkMuDLNeeFgyIeuTri6NyDWEfOq4Xe/ecrrtxcczWsaY2k7S9mYEOPp3CaY3hrPsjIY68kzzTBLmBYp4IP1r+zweEKRPeRJR0phPRgX1nNRiCALwjvw9rFJwTlhoD/4cR7mZRsyJ4qQRZzoqJ828fidE/dLXHPeh2nN6t76sO8ziI/nFUfzalN6cM4zKhI+eGWEc3B7XrGsF/3u0uO8Y5glvHCYMxlk7A1SLk9zIuC14yV/cP0s7JoJ9ri0r8PGkcL0XXmtdZh+jt3hKGecx3TWc7as+KNFzbwylE1H1Vg6GwaYTgcpV6cFV6YDnt8v+PDVCVf3hmSJ5nxV87uvn/D12wuOFzWd8TTGUNYdi7rb1MiVUsF6Z03YqWcxh+OcPNV4F/Iw6qaj8x5nw0/gHGilSCPQsWaQap7bG/Lc4ZDn9weP/Xf4IESQBeHbMK9aThfNxjlRNQbnw/DLSIWarXOhJLF2ThyOcyZbcE7cP3HN4ddBP1GI+rF9ElrdBeE+W9ZUjaE2oT48LlImeW9ruzGjcw63Gaek2BtlPL8/4HCcczDOGWWa2/OaL792wvG8xvaHXBEwypLeKxwcFWVnadtQ7x0WCcNxCl5RtoY/uLFktmpZtYa2c7TWopRinCdcnRZc2x9wba/gA1cmPLc3oEg1Z6uG33n9hK/dmnO6bDA2ZEPPypayd1koBZGOcC4kvCU6YpDHjLKEURaD8hjraVtHYx2dtRjjcUCER2tFkYfQ/71Byt4wY5DFXJkWPH8wYnKBk1dEkAXhPqyajpN5vWmeCOOFwuN7rCNWtcE6xzCP0VE4KNsfhZD4x+2cWNbdtySuhdD3UBeO+tJE24WdZtl2zMqOZd1R1h11FwR3XCQksWJRGW6dlyEis09nG6QxH7k65YVLoxDWnoaGiNeP5rx+tArtz/21Ux0xLZKQ7oZi1VqarsNYT5ZE7A1TlILOer55tOK8bFhWhs7YPpFNUaQxzx0MeN/BgOf2h7x4aci1vSHDPOF8WfHl14957eaCs1VL3RmqpmNerYedQhxDpMKNwBOaQEaDmINhxjCPQYHpPMu2ZVUbms5tEvVUb/Eb5wnjLObSpGBUpGRxhNYRRao5GKZcnoZWan2BU7q3JshKqRz4u0DWX+fnvfc/ta3rCcLjoO4sJ/N641LojKOzjkSHxo66M6wawyCNSfqW2+kgNC/Ej/Ef7oMS15wLbgClgl3MOd8fXllaa5mXHbOqYVm1NMYTK8VokKCJmJUNVWeJFDhAOdgfpXzk2pTLkyL4hvsOwldfOwlPBv20D93veEd5gvJQdZZlZWg7i44gjWPyxGMtHM1rZquaWWVCSlsoX6NVxHP7Q953MODFwxHPHQx5bn/AIIs5XTZ8+RvHfOP2nNNVQ9NaVpXhvGqoOwOECM80VqAiNCE7Y3+YhjjSRINSLOqOk0VoB2+MwzrXB9iHDI1xnmymqwyymCQOLeHDPHwtiRVV6zheNJyXHZ11XJ0UfNf7nv5DvQb4J733S6VUAvw9pdT/cb8JJILwpOms42Res6jvOCfq1m6GZHbGMisb8iR0gEGYYnxpkj9W58SDEtf6jJtNfbg1NrggWkvZdixKEwS3v5Hkacz+UNN2jtNFg/c+dM7FwXP83P6QFw+HjIvQ8ly2Ha8fhfyHedWFmnT/Mwah0nTGsig76v5pIYnDTtM6z6puWZQdJ2VN3Tq8C6WAOIoYFwnX9od85Fo4lHturyBPNUezml//6k1eP14yL1uqNvxMy7qjNTYMbE00gzTuOxo1kyLmcFpwbVyg4whjHKfLhtfmVb8TtptW9TQJg1H3hjEHkwGjPKZIgjc8UqEjcJAFD/WyMnzj9oJF1VE2pnedKIZ5ivfu6RdkH456l/2nSf+/i3NYC8JDYF34B32+CuE/znnqrj8M6p0T87Ih1prpIDgnst45UTxG58T9EteCf9iHcJxI4fAsq5azVXgMr41hWXes1o/xWoWx9/2ki/NVWPcwi+n6csuLhyOePxjifLDqvXlWcr5qOFs0rPqyxCCN+/yJBOcsVeuYlW0YRupCzdZ7qBpP3baclQ2rqsP5MIIp0RFFEXF5OuDDV8e8eDjmcJzhnefWvOTv/M4pb51VYe1Nx6oxoXnGeZJIkSW6b5qJyNOIS+Oca3sDJoNk0yH41qxhUYabR2f95kAvTzSTQcqlUcakrwWniSbunypCB6HGuDCF5bXbi3Bj6/3RsVKkSXB4jPsngsuTZyQPWSmlgd8EPgr8l97737jPez4NfBrg/e9//zaXIwgbnPdBiJahpdj7Oxa2Ig3OiUXV9iHxd5wTh32r8+Pifolr1jlU7xmOopCRfLSo+kCgsANsW0djDIqwG051FJo4OoshtDKrJEzOGOYpL14aMslTVk3H7VnVlzdazpYtjbFkseZgnFMkGryntZ6zVUPVGFoTDjLxEQ6Hs46zsmNRtRjrSeOILEmIteJgnPHByyPefzm0HLed4/as5Le+fpvTZegcDEJsQynCQZZGjLKYNIlIopjxIObyeMCVaUaRxljvWZQt37i95GwZ3BnO+d5mB8M0hN9fmoTciTSOiGNNeJZQKBUO+1a1YVG3LKq2T5GDuJ9msj8K46T2BuHP2Rum7I1yRv3AgIviHadOP5aLKLUH/BLwk977333Q+2TqtLBt3h7+A8G9sHZOKMKBnu89s1EU8h4OxxnTQfpY/nE+KHEtrCcknTVd2OEezWqqzqD7Do+2s3TOkceaPItR3tPZ8P5Q2vAYG3bLlyc5V6YDrA/lF+fDTvJk2TDvpzhPBimDTKMIHXLGhvLFsjF0ncERvMXGeFadYbFqQ3ZFHJFqTRJHDLOYa3sFL1weUcSaunMs65bjecPpsmZWtayqllVrcfbOTrbI4l7Mw4748iTncFSgtQoZHMuGk2XN8bxl1XZ9LdqTxjHDTHNpnHPYWwvjfvq2IkR3NtbRtIaqdZRNu/l7jvW6Fh0FT/Qw5+pezsGoYH+UPtannrfxeKZOPw689+dKqV8Ffgx4oCALwjZZ1R1Hd4X/NJ3F+pBgpvudqLH3OifWB0B3N1l8p9wvcc33ft22c7RdaGoIYUBdGGsfR+gowjhPGnkOxwVx5Fm1lqox/c8S5sg5H9qQDw4yiiyhacPuO9EqHFYuglNjmMU8dzBEKU/ThY68prMsqlACcd6RqAgHm7JI27qNO2FUJL23OOX5/SF5luCd52ReczKvOVu2zKqWsulorSX0pCgGSUQ2TMIU6TTmcJxyZToIJQUd0XaGo0XJ0azmZNFsZtmtU96GecyVccGlSc4gi1Eq6m9APrR11x1VG3zWxoVuwURHJDpiUoQW78NhztW9AZenGfujx1v/fxxs02VxGeh6MS6AHwH+2rauJwgPom4NR3eF/3TG0TlHqiMyndB0hmVtKFLNIAsHdpMi4XCcPxbnxNsT12zf4Vc3hsZYTN+scL6qWdUGHYXaaaw1sQo2rCzRLOuO87Kl6UJ5w3qPd/QNFSEc3vcNZ3Ufgl9WLberDlRwg1xONHVjWFYdjTWsasNs1dHZ4JaIIkXbOWZt2+dOhPyLJFYksWZSpDy3XzDKUzw+/Fwny5A5UXfh57EORfAGD7OYLA5rGw0SLo9DOSDVMVGkqNqOm2crbs0qZmUogQQhDdca5eHg9HAUdsKdc3TWU7aWsqlDl6AL7djrnW8o4SiyVLM/zLk8zXl+b8DBI/4+14l4F8k2d8jPAX+zryNHwMve+7+9xesJu8KTmEByn2u2f/yfuyf8x7qwE421YpAG58S8bEnjaOOcGGQxl8Y52WMIiV8nrs1WoU5bdyHkpjWhNOFRrKpwKGatZ5QnXN0bhF1dHJHFEcY6jhYNb5yu8HgiQhdaOHzTDAZxH1QUxg3liQ6Zv32OQ5HEPLc/pDEmuCD6JpFZ1VI1FkWoo1rrWVUdrfU45zYBQGkvwoejLBxyes9sZXjztGTZz7zrTGiP9l4RRTDINKM8ZZwnTPoxVIMkpkiDNe18WXNjueKo/914BzoOLopBFpwUl8dBSBOtQ/dg5/rhqR1lG9qgEx1RpBHjNAl/X4m+p8nk0iR8/8PgfNhl1114IpiXXV+iCn/PB+OnPKDee/8q8Ce29ecLO8qTmEDytmua2Zuc/m+fY3Yewcd+FOc9TWuI+sxe6xzzqiXuxwFBiGC8NCkeS/btom65dVZxtgolgs64za6RvmFiXraUjSFNFJfHA4osPLanvSVrVrZ8/XQVdsw6+I2NdWgVbiZJyLsMMZRpTJ5qTJ/KZr1nkMZM8phZ2fLG0YKytSzqcE3rPBGhnlybUDdeZ3OgINVhCsi4SMjSGGsdq9Zw0jenlG1H17kgwg50FDHsMyImg4S9Qca4CJ1veRxvxlR94yi4Wdb17liHSSJaR+wNUq5M8xC9qSNWjeX2rKExhqq1NK1F9w6M6SDUerM4lDEujQuuTAsOx/lD//5aY6nbkNcRwoyCZ7psDXXr6IwljSOiSHFlUjz9giy8R3kSE0j6a1oU50w5Y4o3Ef43fpbmAz8MCooswXnPomqJ+rH0SiniSHE4yZkU3/lUiLVD42hWc2tWsqzvdLXZvrUXRfDx9mPth/m6JBJyffM0lE5unlfcPq9o+2YUvKesDXkSMcpDZq+KIEs1oyIJY4z6OX0RMCpSms5ye7bibNmxavtdrHWhXGJDo4v3ftPhZ71HR4phETPM4uDQUCH7uFzUNF1oV257h4dXkMWKJI6ZFjGjPGM6TBjnKcMsRmnFqmw5mlWcrzrmdUfXH6rpSJH0fujJIOXSMONgkuGd4mzVcOOspDVuI9o6CjegK9NwsxykCXuDZNPWPS7Sd6zvr3e/q6blfBW6/UL3XxD7zlo6G2543vtw06k7qs7SdJaPXZvy0kevfMf/fTwKIsjC4+UJTCDxs+vMGHPCPo7wiFqTwOp8U3pY1aHZ4XE4J7wProayDbXno3nFvGy/ZSJHoiOyPGZVh8YN6x2DLCFPw85wlCekseJkUfPajTPOlm3fVaYwxtJ0nmEWMxkkOK82ORNhh6hDSE5n8R7yVHMyq3jr5jmzZUfZtzJ3XahR08dwxlGYT2f6Lr8kjhhlmjiKUZHCO0/j1w4FS2eCP1f3UzuyRDMqUkZ5wsEoZ5iHqdGdNaxqy9lqxbxq+wNHv8li1pEiiyPG/YDSYZZscpZfu13SmVBW8ijiODhExnnCdJgxymOmw5T9YcgIyRL9bX9nrbEsq9BqPSubTaBS3YV5f511mP4GoZSibkMjSNkYqvXuPQoNNIM8udDmCRFk4fFywRNIFlXHyegH6JYnALRoHBEJHXp0td8BhRFE6wOd/e/AORF2iab/xx06whZ1x+qu3XCk1GaOXIjqrDk/r9Ba9cIVHuX3hxmt9dw4WXL9eMWyNX0o0J1H81GWEsUq+Hy15tI0HGzFkaI2LniklWJZt7xxtOLmeblxjXTW4Sw4PIlWZEnY8Trv6Vw49Uu1Jk5B69A67JSnbcJcOWtdnw8RPNl5qimSO/XgItboRNP0Afy35xVVbfoBpyFJDjxah+khe8OU/T66s+tF+KwMHm8PRISGlnFfq54Og/94b5gyLsKu+0GHcbZvHZ+VLfOqZVH1TovOYExwsHTGYXuRxXvazlAZy7I2VLXZBNPrKHTueedREX30Z8GHroy/4/8+HxURZOHxckETSMrGcNw/TvMn/3W6X/tPsdYSY0jpaPSI1ff9RXKtKHrnxLh3TiQPcdLeGtu30IaP6zbmpjMsqlBrjKMQ7RjHqi8dhByIbx5X1K3p/bUpgyyE04/zlPOy5nffOOXWeZiO7H34M6115KlmOsx7vzAcDFKePxhuIj1nq4aTVcuqC4dqN06XrGoTBoJ6aKwPVrkoosg0qVYYB1VncNaBChawJNWEPpLQ0dcZi/Kq30VH5FlMpiOKPOZgmDMuYvJY94lulmXbUS0qqtaxqBoa40N8JhDHYcDoKE+ZjDKKJKLrv14tGohCzGWRxORJtBnZNB2kDPPQGTfM40279Lf+3nvxLTvm1To4yNBaj+lLD511RLDJQ/be0dSGoybY+0IgU0QcQZHHm6cARbhRJakiijTPHwx48dL4sTYCvRMiyMLjZcsTSJrObkbRQ9ghdR/4YfQ/DvkrP0u3PGExfD/JD/wEkz/2o0BoBT6c5OTfxjnRmlAnrfqA83XTyJqyCdm9SinyftcHoT5ZNobrJ6uQP+E8eRZzua9LP7c/QEeK68dLXn39hNNlQ2ccrTF0XdiJjfKUpAiOCh0p3n845IWDEbGOWFYt148XnCxbTpc1R/NqEzjkPRjncdaidcQ0T8lT3R9OWeaVBe9Jk5g0icMTgQp1ZOtC+7PHh642pcjicJi3N0wZJJosjfEEd8rqrqeCednRtBbrQ8kj1VE/QSRlVMQUsSZSoauv7effxToiSyKKJGE8CJ110yLtp44EEX67J7jpQn7IWnwXfVhR19fB14elYVSUIotDi7l3Ie6zbJp7RkYpQkJcZx3OGXykiONo01jSGUdjHMvG4X3L5b2USE22NvfwflxIp97DIp16woPo+rFJ66Gd67HxURTsUta53sMbMigghKZf6uuVb8dY1/+jNf1Az2/9d6B68eqMJ9KhycD7kDVsrOe8bDiZN5R9y/Wwn6ZxOM64MsmoOsfXb81543i18Q83XSgH5EkIQUcpvIfDSc7z0wGjItQsz5YNt2Yl52XD+bLlbBlqoaGFOQTAR0oxymMGqe538pa6c0TKo2NN0rswItg88oeKhe8jLKONCK9LA1kc4VU/V64LLoTzVcuiMRgTnCJpEpFoTZZGFHHIvZgMMvJEo6Kwi3YulExSrUmTiGkRcoZHRcI4D+OfijTelI06G8ow87LtXQ9hBp6xNuywjaM14Sawrs8nsUYraPsbXNV6qrbbtMJ76IPqw40JQvkmTyKcClY/az3GOfzZN1E3XyXqFowTRfKhH+T7/9F/jB/4yGM7zNudTj1B+E6xznO6rDfhP977zUDKPI03zomQOXGXc6LPnFg/9gYngt0c3twdabkmUgQ7VRLRmvD+KIrI0iBinQk76NM+9L1zniQKYnMwyri6XzDJE44XNX//6yfcPC05KxvaNtR2IURGpjrqd20RB6OMw3FOkWq893zzeMnNfsTSomw5Lxs66zcCEwFJHDGOY6zrW7DLrk+CC7XQONIhz0GrcFNxhNqx7yd7xCFBbVgkTPpdddRP+ljVhlnVMFuGxhHnQut0EiuKIibWmiKN2B8VXJ7kjHKNc4pVEw4RcZ4iDTGW4zzsgPeHIahnmAdXSGeDK+TN0+Um47js68+d9Zvc5M6GDOMkUqQ6CutUoTW6NSFoqWws1gfRNs5jTD/xxIXfXZqEg8j1TMHWGGalxXkDKGIN6fybJG/+PTK7IsPwYvcm06+9wnMftPCRf/4i/jPfIIIsbJ/voFHE+VCPPV3WG/dCa4KjIO0Pqb7VORFC4veGWQiTacwmSvHu4Z9rFEEgB2lMkcWovvPsdNlu3mNdeDQ+W7WcrWrKOuQ7ZHHE5WHG1b3ggY2U4sbpit/+2jE3zlbh8K+zoWnDexKtUX3dVmvNpXEQ4kEaU7Yd/+CtOUfzkqqzmwCezjmU8zivQPl+xxtKG2dt068wuBiSOCbRnliHSSbGhoCg0PWmyOOwkx71zRqDJIYIVlXHzfMqiFtrMN4TsXZVxMRRSLcb5SnX9guu7Q0okpiyMyzLjnkZOgLz3hsdmjNiDscF00FKmkQYG26aR7crFmVH1QXXQ3A8BBFuez901DejxFqRpwlJFGH7m+H6dxnm57m+/u1oN2WJUBpJ0whNiAXtrKde52BEoXFm1Ds1skRTJJrsj75IYo8oqBmz5HlucWBnZL/+78OfEkEWniUesVFkPUH5ZFFvygidCdOH036nU7WGztzrnJgMEgappuk814+X94T2rAnJaJoiDX7btX2qbAwnd9WlAYy1m9fXQeWKMIb+YByGbo6KhLZzfO3mnK/dmnHrvKZs2k2jRaTYTGpeC8HhKGd/mBFFYc7dH745CxGUfVt06JZzmx1xHEUkMSgijAtJa3iHihRJFKF1qIsqpcIjuPN4Fby+WkcM0hAbuj9MiXVE3RnmZcdbdbnJNfa9qEeRIlOaVMMgTTmcpP8/e28adGl63vX97uVZz/Iu3T09u0aSJWwQBmzZxjYYU0Y2TuICE3CFD6FSSfCXVCgUhxQVEkIpJEXFJMSEpCoOpIiLhJRDWBLjKrAV2wrg2MjYlowlW9Y6S2/verZnu5d8uO7ndPdMd0/PTPf0TPv8q1Sjfvvtc95zzvtcz3Vf13/h6YMpT+xXWA2LRpgUq2YQU6G07MqMKB1Hcx7vI8u258bVhqbzNIN0vyPlrO89fXK0s0Ze3yS9RzJ7Flph2ztOO58k2UE+9yHI+5BGNsbI7Duq8eYpikijoMhEsl1l8lnnVm4yWVo8HkxkmTj9+P/DnCUVLSZxZjwKt7hC1w6YJL55O7AryDs8XLwBoci6HThatttuVpYvQr/KrJHOqPcUuaGoMvpBjGuKXMviZzPwapTJhHykb40Lmhgjq9ZxumpvK97jIunGoqPphpS7BvMq5+mDmicPKoxWnCx7fu03rvO5qwuOVx1D8nAIETKjqAvLpMyTP0bG4VTGGi5Erp1tODrv2PQDJ8uW64uGrg+4KOwApUdHMikPPgWYGqXFb0IZYlQoLTNhheTp5bkmU4ppZTmYFUzznBADy9bx0slalmApxmhkjSglhb0uLPPEEb68X7M/LVBENp34BrsQyKx0lWQGpSKzImNWie3muvO8crIRwUWfKGeJJ907D0robZnR4jGRF5LunHyeh0HUk5vOJdn3TfOkECI+jos5OQlpLbP3wQW8imSZpi4yUQdmNo1sNLk17FUZB9OCw1m59ccwRk4agw+42QU2S885cxwWhyWgYXoZTjdisL8ryDvchkfhD/EgcB9CkXbwHC0aml7Mf3yIonBLx+XRJD6mdfnpukcR2atzrNXcOg4uUsJEVUgn/OoNuQ9yfD5d3e64tm4Hri9bzledzFuR2emFacF7Lk7Zq3NWrePXXznjC9eWvHy0EjtJIiF1z2VK16gKcTObVSIHnpQ5y6bni9eWHC9bjlct1883nG0GvAOtRUYs4gmDtUrkcEiXbZTGB08fIlppMq2kM9ZaOmgjNprTMpP3wwXO146TVS9mQyn0VJZdpGVgJoyKac6FSbkVnORGs+5cKsKpW0xLTB88dSFLOqtlzv7i8Yp2CAxOxhC33tyMFmOkeZWRZ4bC6u2idNU6ztaScLJqHZt+oHMRl5ZwLghzIkS54xjkRqV13N4YysxSl0Y41UZTZYa9Sc5+lTOf5OzVGXlmCUGWuMOv/kOWP/s/cro6kmL7jX8SPvAR+Ib/ED7xg+C6m78otpC/f5uxK8jvBjwKf4gHhXsIW69e8AAAIABJREFURXonlpBjinKIMaUIy5G7HTxH52t6L1aKcoRXHExEZACypKrTHLi8ZWv/arzacQ2kuzpbd1xfbGg6McgBKepP7onNow+Bl4/X/MLnj7hytpFlng/pZ43kRjOfyNx6kjrxg1nOpMgxCq6dN1ve8fGi5XTT0w+StmyMzMONNdRWUVhLVEgBCSJgcEE8gKXgyOsbO7+6sBR5YhoMcRs/NfKYQxC2w+h3MS0ts1J40bPkonaQsgA3nYwjrja9LMi8jAZ8EF+MKqVsbzrH2UrGAmPKRlKGU+aWw2kmjmuZjAYk/Tqw6QaunfW0vXglD16656b3UoRjJIYAUUm3b4QhUqfim1lDWRgKKwtL+dxz9irLrM6ZlTlFJnPj8fNdth7aFAz7uZ+AT/zlm0V3dQ0+8YNoBdnXfBc2i9h/9t+SLV/Ezi+TffsPkP3OP5KMm94+7Ghv7wb8lQ/dpag9Bx99h9tLv/pmAng74fgjP8T5898JJObEZ3+C4Rd+hLg+pamf4uy3/BvEp79eTNiVHFP3JwX7k5xJIZSpujCve8GMjmvLW6KR1l3PjWSe7hKDwYdIkRku70ne2yKJD87WXUq6EGN2n7LiJrnhYFoyL6ULE9tHGU+su4GXjldcPVlzlGbQbT8QUzinNoitZikx9UYrhjQj7ZJSLozS61SMbjrAGfFIVmJvqRLFy6VOGGS2ao2mzAyTMkvdpHSPl/eqZOCDLCpX/fa1jc+ttQhdrBEtXTfIYsxHtTVIKjNDmWkyKwKPMrNbJkrvJWNv0cjsuEtjjKYXQ5/OB7yTObm1Srp+o8gyS2HkZzdaHr9InXVV2CQcEdHJtBTGx+shS4b09m9+N9nyK1hEPJSl4YTee/au19CYq2jSe/kWsaO9PTZ4BP4QDwy3CEXC+cuczr6a02/4AeLzH2FwSen1+X+C/+SPELyn4SJx4yk/9bfQFrIXvpkn9iueOqiZFNl9+9muu4GzVb9Njw5RpMzXz1vONh0umcn4EJlXGU8f1CgUp8uOVScsh7N1R9s7huTHUOaW/VpEDJkVz10pHiI2ODppuXq65mTZbeehgx+XT8LdLa1mVmcUmaVLx/Vu8MnCMhVhayhyg05pznZL35LHiAh/1jsRkhijpZBpiS4qrHB0p4XhcCo2lnVhGYbI0bLh6ulm+7wyUxaRidYaoyQ5epHM4VViaUzKnCIzWA15bjHJC2NwQmE7WrRs+oE+3VjWrd8a+IgiUX5WndKyq8JgtcyQc6PJMul+J7mhLrPU9cp7PS2y7UL3VowMEkkB0an4puKejP23WH6aO0Z6pmsoxkjnAm0v7Jg2Se5BfKTLvZ3b2w4j3mZ/iAeN+Nv/GOfv/0NcP29Yd0nxdX1JCAFtNPrTf5/eaxwZNQ01LaVvufyZH+Lyd3/vfR8bR4bG6brbLgbXbc8rJxuunW1onE8KNSl8o4FN03uunzc4H0Ua3TqaFLqptWGvkFmoNSYtokSUsu6E9bBqHWerjrNNx7ITJkCMAWsMda6w1jArLHWR4WNguXEcnXf0aaQgPN+RBXDThrMqbKKTqaTIk8KZW01ZyXzcx4hGk1spcvOq2FpnZkbT9J6rJw2rTpRuwnbwW4WbUdJ5Z0bmu0bLOCTPtBS89PNYneawLrBY96y7QWwxh+Qh3IxdsMjAR2itsEqjjEk3pUTBKzMmhWFa50zTMm5aZkxSKGmexjI2Fdyx2Nok0JHk6DegoHvVNeQwtBQ0s/fRHq22ir7b8LmfgJ//YcLqi7BXvC17m11BfjfgbfKHeNBwPnBj0fLKyZpVO+DSsk66MSlUwQfC5oQDVszZkOGYsOYip+QrJ2qH14EPkfONBJZKkRCfiysnDedNd5snwiSX/LbcGnEF67zQpXxg1Qws20RxSxJpaxQhCuNjpG1prVL2nuPGuXB4u2Rgo1FkmcYiXhDTQkQS55ueF0/XuEGO6korjNYUhaLMRCFXJDpeObqZRbbfdziRQjWOFrRSVIWlSv+uLjRFZiElklw7a1i2jmWTkjwGJ8wMFcmtLDz3cgkWHYubUSrFKcnrE5+NwGLdse6GrRvaph87X08/eLxPk2Qtqc3G6m33mmeWOhOT+/1JQZGimOZlxqwuku+FUNPG7nbseB8UQox0v+/P0/74f0LrIi0FDiuLu2/4KKQkGZCbY5lbys/9GOUn/gyFO0cT4Zy3ZW+zK8jvBjxkf4gHhVvVcKerjuuLZtuphiBUq9HfNjOyQa+KkmrqYbWgoOMSx1Skxcvec/d8vnU7cO1clmWt82y6gZOVFOZ+8DJj3c6fcw7qEqWlCwwpy27Vdlt7zN55lFaoCM4JD9aqVGi0pDrHAKcpbqhzspDyMSnojGz+J6WhMIp177m2aGh7n2avIknOM7OdgddFRpUb8kykzigZPYwGPaPXcjcEPH4rdR67xDwzmMRBPmpaTpYtJ+uOtg9betjoNVHlUuyt1WjUlvkwelx0ved01bJqhyTD9lv+cJ+sK6WzZqy/WGvIrBRzo0WlNy0yDuclh5OcWS0ubtK5Z8wqy7TK7jvJ482gT1zktpfPsHMBnvou+DYNP/8/yUJvehn1jX+S4rf+wWSHKrTIrfHUP/0YuLPbH/hh+3qzW+rt8BbgQ6Tt3S0dkxDzT9f9Nr+OGJPqSpYyRkngpoIUVKnIPv+PufjT/xFTd3LzwbMKvuevbn/5xwVLO0j6xPUx/ieZ+xwvWpaN+Alba7BafAv2E/k/pjTipvdJWOI5b3oWa5HfDt5v1WLWyojCKEVAKFdjmsSqcULz8gEfwtZwvU4jjcF5Vum9IHXMKEVmocyE5TCpLFUmzJBx1pllYnJTGIMLYuLTDUFMeRKrQGuZ4WZWb43UhX/dJ9OduJ3FWiNFclyKFSmXTyvhLnsXWKXXs0lCm0gkJLe0zo2m7RKxhFYYFCYVX538kedlznxScHlecjiXvLzpV36G6p//d1TLL1HNL1B8x59F/46HU8T8L/8o3cf/Es3ihHb2Htpv/PcJX/WRO35vlpZzZW4oc5toeHcZe/yFfe44c0bBXzi7w9dfF7ul3g4PFiF1a00nx9bulqOe84HzbSyRcE/HYpJZsy2cfYRJKfPPrUn8t/5R1DxuTwBh/hz9t/+ntO//Q7Snoigb/Q8WiREQgtDYjlctzoloYVpnGGTUUKdjsNKSfNH0A96L4u/q2YarZ5tkwC7mMuPFOqttooyBEQoDJ42kQMuCTwQKmVaUpSjLfIgsmgHn+pveyFptFV6zMmOvFm5sXVisUmSZBJjWhcxWN51jse5T+KpQvOoi2zInOh8YOhnHDM7TDeE2mbMxGmvVzWToUpgVIark9zFw7axNqR/JGyOIP/IoXx4Si0RrhRlfoxXfaGuFIz2f5BxOCi7tVVyYSrp1lTr+MjdUv/4PKP7fW8ZrizP4sT8l5egBdJajK1/bO9pf+Yf0P/1fgwvAPizP4Wd+UHw9PvgRkUanU0GZvz4j5zY8or3NrkPe4c741I8Sf/JjtIsbbGbvY/O7/wPa933na74thLidJRaZHL1Dkg5nVi6ATecISUJs0vx19JwwWm3zzdrhliPm+PhJtLFshuRNIEq6ZePRxK0wJBIprGZe5ZTJS7cbxC0sRImo/9L1JSdrkTaP3U9uFHWRYYzaLrFyq1mse24sG5ZNjwsigdZaRBqKuOULi8RZbeXHJt2E9pLr2/6klFm0VShkRponqW/Xe9btQFRQWkNuDS4GmsbRe4myV4gizQfP4Emd+Sh1FmaEzIQ1dSYKtBCRG1BIzm4hbl3SvJdCLjJreXytpZCDMCDUyPbIbUoGKbg0L9mflEn1KIVuLMK3+Us/QIrmraeiMfHjNlO+//X7ZPwA5PSUdPK/+UXyj/7iW0uMvgNd89WntjeIXYf82OEhq/W20US/9Pdo/vFfpHGayFOwbOCn/jJ4BR/4CEXi3A5Out7x4vRBjvE2zW7HyJw6v+k5MYZhuhC5erp57UWWMF6M3SCKvRgDZyvZ7o+RQoOPhCEwLUX2mycubNs7Vq2MIl4523DtZEPrvBQ3AqAoU6ZbmVmUEgVcOziuna/F6cyJ01lMhVvFCJjETACiIiALSqOF4TArMy7NSw5nFfn2OCzuc+OfhxA4T+50CjFrX7Q9N7pG1ImIPwVKSfFMjmcgFppVbqgzgxkpd0ZvkzdaF1A+EMP4Kscbj3TSHrDaSISTjknqHbdjlyqZD12aV9u4pGmV3SY9L7O7i2+AN03RvBft7NXQaeFarn6NkpbyFg8KABYreKsexo9ob7MryO8WPCS1nkTT35wDR4BP/HVwN7uenJ7KnVN/8i9R/Z5/nXU3cLxot1ltEvkjx/RM663Byyhk2HROVFe5Zd151p1/zc+hlcxYtYa29/ggXdui6blxvkndsRSnkEQch5OcSZknBdnAjWUjWWobETycbZL8tz2D9Q0Kv2ZmHHtPvR91+LyIMQafIoA6Nu3AEJJBDxGduLk6WVqiIjEoXAhYhXg5THL2qox5lVNkNh33hRkxKS0aReOkq9/0butSNkbbk0YMWbqJuRBohoDW48zTboUgJMHHJBeFnkbhonTMkv6c/I5DxMdA76JENwFDulmGdAPJjWZvMubW5RxMCuYpRmmavIqrNHd+Q53mfR713VaplyTXd6KdJUgDMN4QzE0j+70pnJ++7nO9aXzt973ti/NdQX634AGlOY/RRKOV4Z2602z1MjUbKlpqGkwSFa8XLS8f33RS86M5QgrPbNKM1wcRQww+kFmJeH+1W9aoHhvne85HTteSDnGyFgHH+brbGsAMXvLhblplykX9ysmaRdvT95KScbbuWbbyMxityLpzJutfZ+7PsQx0vuTkyufo17DK9lg3sjzzIRK2RTi5rBnpREM64mulUodfyYw30xS5zMptOjWUmWUIjsV64KXjNe3g5OTgRWrtka7YGrUdY8hoU2hiVS4jk8waFNIV17lJEmmVbDUTdc4qCjTew4AIMlwQMx6RPctnpJGU6kmRb7Pr9uqceSXz4El1swC/OrXjDeMOFM1ga6GdrbrtWOpOgQCQThu3zH2LOwhC7vVc7wY66L2wK8ivh3eKqc+bPAq+XjQRCPeyKuzWE8LuAefH279vyTnmkM30BXBh6znRDbIkawfPYt0R00JNTOLZphLftt1Oc+bRaHzVOq6ebjhLsfFHy4Z164VS5sVmsjCaIksZcD6wWQqtbYwqajrP+Sa5tmlFrjVFkWS3x5/B+Z4z5ulwm9P4inbRE0wrRRjx8smUIbdys4hxnNeKP++sNMKIUFAWIkmepJuDNZq2H7h62mx5zDFZRHoi3t0cC0wTC0LYEJYyU9uATZ04xwqZ945LqEAkRokbyozCpnSU81WX5udy5A9h7IqlAM8y4ftOq4x5nTOvxCx+DBEVscsD9mr42u+j99D+1A/SLo5pZ++h+4Z/D576CCzb275VkW7Md6Kd3edzAQ/2+nzE1/tuqXcvPPjB/pvHfS5LBh9uG0HcqRMZucBVcZeuKL3ufhg44YBzZgxmQvu7/zTts98qRvGJRtYNnpBi6HVK63hiT9Ikxrnjq7fbo+Pa8bLlaNFwbdFyupQj/ahyE7OzQAhiNjMpMkIIyXReinE7CPe4d1HmtJnwa5WSYtYOgc3Lv8JAxoDd/jemqWu00zSLFYmyQuPTCEDsLzW5tWijqIwEcs6qnNwqhiGw6geRCN9y0gjISCWC8H/zTGhuuZXIozzDACi1HXNGFC6xH/QtLJHcakKINJ1j2Q2sOydiESXpHz6S/DUCRTIbmpViiVmXGXtVzsG04GBSUJejQdCDzYfzIW7VeqNXRbhLTXlDtLNHgYd7vd/XC90V5HvhnWTqc5dfFvev/lU2v+UP06SE5DtHE6ntdrwuXhsmeStGIcKVn//7HP3s/063WeDqy/iv+cPw3DehUwfTDS75QIgpeWE1T+zVPLFfva7j2rWzhmvna66fNZw1A2030KaOOEZhNBSF2FfmRrFuHatuEFGCh00/sEnKP6PFFzh6MWZ3XhIkRBUYic0RvVc45OeWGKSIMWCKPawWb4gQwFiV5rkKrQx1oZkWOfPaorTGpVy3Te9wPrmKJW/hkMQSdS4Lxr00BsiTCs9olcx3RFghMmSfFnUSp1Tckg24aMQVDmSEEpMCb+RAl5lIq/fKgmllk8+GFODDacG0yh9KwbuNdjb4OyaxgOwE3hLt7FHg4V7vj5ZloZR6DvgR4EkgAD8cY/yhh/V8DwXvJFOfUSDxk3+RZnHEZvZeNt/0pxkufyec3T5b1ilvrs5lDFHcw6lqXK60vWfTD9xYtCzWPWH/W+C7v4UQ5IITkxkNyKLocDphkoQdsyrjwqy863GzGzwnq5aXjte8fLLmxvmGs7VwlvvECjDJfnJaGeo8Y/CBdTNwHsUvYtN7Vq2j7Rw+eJQG7yKNZ9vVklKUQ2RLdwvUKNOifEzMhogxAW0naGPSn1OShFLkufgp700yjNb0gxOhxygGSWbxo+1kbjTTOme/zpnXGZk2VIWMZUhjj3XncJ2MEmIUnq8wPDRaa1bNwGky5t/aWSq2QhYfxIB9mmdcmJfJatJISsck52BaspcMix4kXpd2dgvEclM6X1m8vcO63/vBO+B6f5gzZAf8QIzxXyilZsAvKKV+Isb4qw/xOR8s3gGmPrep4Z76bro//l2v+R6FhHNWhRSTu23GhRImF9Z4ofmUILHuxDB8vOBUjOSZJs9SmjCS05Znetth17l46t6t4K87kTZ/8eqCL1w75yh5H9/KLigyoVuVWUaZSYrDqu1pusCyHVi1SYXmRg5tSEwCtvHu47O7AN67NFNN+0aTowJo05L5XvjG+QxbTNFG2AbW6pSObJJEOnK+GXCebcqzSou4Usssea+Spdiszm7aZKZU6mUrBu8xyghDoShyWdRZRRq9DCybJNBIp9TBh+2R36b07Iuzimmdba00DybSAR9Mi7e+gLsFb4p2dsv8955UuHcL3gHX+0MryDHGK8CV9P+XSqnPAM8A756C/Ai2uPdSw92KMZqovtWM5hZIMm/YbrXvdrzcdI7zdbed00pGmU6uZmIw3nQeYxTTlKFWWM2FecmkyF7zeDFGFpueT33pmM+8fMqVs4ZNck6LCMMgT1Le0mrq0kKAdd9z5VRsL8UxDAbv8T4wBI9SOo0IEFaHkq49hIhXQlXzLqC1lkIYQ2JLKLKiQuuJSH5VUrbp5CSmRTU3up35IDNxrRV1IRxfozVFrpmVRergZRSRG1nChQDt4Fi2PSGIMZHW8rNVRj6b3nu6zrFOpu+RSJcocDHK+zKrhMc8KbOtKfz+REx5DmcF5QPsgN8U7SwV4Ad5I3hH4R3A2nhbWBZKqReA3wX83B3+7vuB7wd4/vnn344f5/7xNpDDR0+CkYY2xhi9Gq8XTTR2V+3gUkDk3S+wsfis2wFrFE8e1Ld9r1Yy71y1vVC9SpuYE4oLc8klezXONx2ffemEX/rSCS8erROlLqYEhyhy4FwxKWTmTAyse8/Vq8ttKrTzia7lAz4GYpCiG2NKC1EKm2lUFJWZjxIr5KMiRJnNeh/kppLJTUW4uwadxB8meVzkdjxaq9TdicHO6NOgFBJ5n8JJq5SAUeeGsrBbH+BF06VcO+mEVfpMY5Tn2wwe7+Wz6F28ORdWiro0XNqTG9u0yERmXYkybn8i79ODOPaHGLe/E/dDOytvkULfk3b2uOEdYOL10Jd6Sqkp8DPAfxFj/Lv3+t533FLvIWCrhktdcNO7OxbOMSXhNdFEn/pR/E/+53SLG7SzF2i/5Qdo3/ddMkP93E/c5mbF878b/ZV/Rrn6CuXskPLbP4r+7X+E01UnG/uEkC5OCY6UuWWIkboYPSeEwjatxFth9K04WrRcOV3xa6+c88rJZvuYo1eC1TCvxZ/Xak3rwjZuftMHgneEIHzlSExjhrCVIo9vjEk+uCS63Rgx5MeoIka3NeH2bilkJKqY1ZTJ9L2yhrrKyI1CaS3GOUo4wqT5rrixZRxM8iTFlg66c55l09MNctIYjeJHYx+jNIGwjVAaAzrHuXFVZMzLnLqU+W+RTjmHU7nJ3U8Cyv3gNrezwd/1lPWWaWc7vBE8epaFUioDfgz4RzHG/+b1vv9xLchd6oA3vbuNInUrMqO3I4jqlo30ONvrBk/7qX9A+5P/Jb275QFsAd/2ZwBQn/ivKNxiq+kv6MiRIukwnNgnOf+2j0mwIzLLJJn8xBhZdTLfHS9KF4IYhxc2zZgl7ud8M3C8arhx3nK+Gb0hRCWWaZhU2Xbht+kGms7TefG7kLmssCV8lFGEQoQfYyEzWgry6HshfsQ+zZFHWbBAaynYZXJNy2+JOcqNLNgmpWWvzClLQ/DQDCMTQ7pyY2RcczAtuDgrmeQGmxKfB+cZ0ujHpXiiJn2OMSaFbuogx6JOsr0sM/E3HqXdVSp6+5OcaZkzKe/NeLkfPFa0s8cbj5xloYC/AXzmforx44RRDdf08t87XSBZ8iUYu+CRoC+uZp526F472/snfx1uKcYZPaVbUn7yBynpKdwXXvOpexRn7HHMPoPL8D/3v+Ce+X0MSYLstp4RbpsOATKjRcHJqmPVDFtv3JNlw9Gqo23FOzi3mjoXBkaembSZDxwtxFlsSEo7cVBLhjZa4uxx0HtH8BGlSZn2Mi5REQYf6WMkuJuJyVsfXmOoCkWR8uwKa8mtosxl/jpJqRlaqeRt4TjbDNv3ssw0h9OCC7OSw7QgszqZ6ygZI63agcVGEpH7IdlRhoBOhjwgowdSzt4s8Y1nKYJolB7PKknFuNu46Y3+bj22tLMdHuoM+VuBfxP4tFLql9LX/uMY448/xOd8JHhDargkyMiM3nY3i6a/jfVwJxitKFdfpKSloL/NUCUujreRNAMWh2XAcso+RxwwkBOS1Yxae+Lf+SjqQ9+Lf+YbxXPXCoNg2Uga8Jh7Bmn50w8cffmzLF75DXAN0yxj/sTXMEyeZXBJreccw7Kjc367qBrj230SMchIQN6vtWQpCcNAKXSQsEtiZPBpSRfFo2FsQjNLiv6xcsROHWhm9VaKbZWm83Iju9LK6WD0Bp7VklJ8OJXY+2r0iTCK4AKnTc/xouVs3eO9RNqP7mrjbLnMDCp5doyMkzH9oi7MtgjWuaUub7/ZvlG8EdrZeIN/V9POdnioLIt/wn226e82uOTNey81nPge3FTDyQxSaEXHy/aetKJXz/YyK9t+N58zLDY0lCyZbouvmz4t/ypZEa4pOWcuMTWQrCFFJKyImOYGm3/xt+l7iE/9Ls5T5zhSubp0RF9tOhbNwPLqbxCv/CKF96wpWQ0Wf/VLqI1iKPdp+oDzqWM2Gqt1sngMDCEQkohiCJJpp9VNu8exWwwh4qIwHKK/SWnLMzFZnxQ3E4dnlcVasy2UIKY8Z5sepUYqm6HIIMs0e3XBpZlE3teF3YaSLtuBK0drTlf9dhnqkkG78+KxXBVma41pjYxERhe0SSHR81sXtsIyKbJ78r7vhh3tbAfYeVncF26NJpIk4Tup4UixPEkme8tFee1sw7IZ7ri8i8n+UNKL1TaZ1wWxtlxs+ptd0Yf/LHziB8F1Nx/AFvCN3w9A+zM/xMJnBAwZLnXRHouMPVZUnHLAkgnRa9Sv/TT64tcyqwu0gn4IDE3PJkX5nG86oWZd+RIrfwFHtjVz7H2GW6zQYSJBnDZjcHJK6LYGC2nxJWNVNGJ8vuU6C7GBED3ej++jwmQig66LjEmRMa8zpmVGphV9iFLUk+ey5MtBmUIyFZBZw0Fd8MR+yYV5idWabvCsOzH8OV/3bPqeZSvqQJdy6sTYxnAwzZnkFp3CRieluLlNq4zcGvGlMOLmNp543ugY4oG4ne3w2GFXkO8AH+K2+I6UrFdDwW0jiLstSEbbyE0nTlxEWUKZlMqgtfBr7zYLvA0f+AiZjtif+++xy5fIZpew3/ZRwtd8D+dJ1PHEPxeWxVBfZvH0d3D68mdZNj1nTIkoCoatYs22R9SVRSMChvNNy8vHGxaN5MU13UA7BKKfJydhTcCIMTwthXcM6kkWm2HrDUy8qTKLSAadSm5mgYhBkVmZ044pFaRg0FGkcXMOm1HnGTKmjaA0mY4MMRBAZrRWfoXzTHNQlzyxXzKrcnyIrNqBF49WtJ2jdV7iijqJqu/TaMYa4f8+WYp0uchkrj9L+W9FZlMun5x46uLmyOl+8UZpZ/ftdrbDY4edlwU3xRib7mY23KtxJzUcgEs0sPGYe+uRVwzA2Wa15fbeabpjUu+Yuptt/3vnJN7BB46XLctGfB4Wm55F07NMS7iIhHX6n/thsvZknDhT01DRovN9jr/1z/GVo5VEGnWObnPO0CyI3oOxqGxGGFbgAxYHBELqu6MxhOIQEk1tpHfFcV6sFMaQMtnkdTjnxZvXgzGKzEjOm7WGSWGTCq2kzo0EkQ4BpaQrjlEKb2HtVuk3LuaKXNN2kkU3DAGPCFpW7cC6HbYdsDWaMnXfZWaYVZkEc6Yl4K3y41F8MymzN8RI2NHOdrgDdokhd8MoIR6LcHuXCyZPirU8mcP4IMf6ddveluTwesit2DpmiU97a7G1Rm3///12Qj5Erp9vePF4zWLTs2qk6Gy3X4x0LU+ZW/L3fyvhs/8Q6xssA+fM+FXz1Vypvo7VL7+YvHoDYWiIbo3yCo1F+QicYUwFbHDeEBB1XjCaaGqZSacOEmWIKqTfPL2lhWUZBC/G8zFGtBmVgIo6N1ycFVw+mLJf5xilWHeO3jkRfKAgpsVcUhPuTcTPVynFuh04WjRiXB+TM1ojkfUoee/rImN/IvFMs1JGIJOiYF7fPu+9nXp4f3PZN007y96E+fsOjz1+UxTkO6nhYlKRjVxUHwJW3yySkvUW6F1g3b3+c7zR7vaNwofA8arlpaM1Lx2vWH3pF4mf/yliu0CVc8JQHkg7AAAgAElEQVR7fz/uid+2DeaUBZRh2fQ05fvpn/kejl/+PDd8RWOmuGyP4Cri4CUdA4V1K6zv05+FMofXeAaiPSCyJviANgab19isAm3wPkCKWVJxVLmprTnPppfsNulQLZlV7Nc5Lzwx5cn9CWVmWLaO03WXXOTYFvMiM0wLMVUvc5NsJyMnq46QxkFN51h3bmubWWaWaZWnub7ZGrK/2v/h1rn/67ngjXijtLMys4n9sKOd7fD6eCwL8rixHkUJd1LDyXG/23YrdZ5vuaV3woPqbu8XPgTO1h2nq57TVcf1RSMLPoCrn8Z99scJPqLR5O0J+a//XToiZ9MPsGgGFmk5N3jPunV0wwRffIixfIzNdGGFFaE0DJuOkOS/YyabJxJ8QFUVOp+QaTHaISKrQi8ZeintnlxrYgxs+kATHEqxZYpMy4xnDiZ81VNz9uqCxabj5ZMNy3YghIhK447c6q2goswtNkmbu94nWl3YdvVKy43wYFpsbwKj3PnivGRa3i7zFsbG3T1AXv0Z/KZyO9vhkeOxKcj3FU2UjqTCgtDkdgU8/O72fuBDYLEZOF21nCRFXIjQ9eKN2w4yk9YKss99nMKviUSO2ecaz3PuZ3S/foP+4ADvwUWPS967Y/EVObEc463VKBSDd7gYwUnnq71jwCCZzgKdjIYkzUKYE8YoinQDk45Y0w+eVRLCiHeFxBodTgre/+ScS/OSde955aTh0y+eEry4mskuUFFaxUGdMymFzRCjSJpHf48Qw7bzx0Zym2PSc5e5Za+WTvjWuCirVeqAs0R5u3OR3NHOdngn4LEpyGfrnvNNf9vX7qaGG/GeS9OH0t3eD2KMnG96TlYdZ+uO85He9pX/D/8r/4B103JWPsPw3LehnvggWmmC95z3A8f9Eyx5P2tqHJqIxmPwXkPyk9BaFmZ5opZZK1FIPkS6FL+k0jJOK4WPit7uE1iCF0GHxmNMxFZ72FwLy8EYrIXgAuMjNM6zaXpIaRelFQOeJ/crntqrQStOVy1furGid146R5WCOY3hYFpI7l5mtwvRTe+26RlFIZJnib0XUUaWKYzS7KUCPna7I/tl8jpjiB3tbId3Ih6bglwXllU7bHPh7oea9HZeWDFGFk3PyfJmARbrA5mDDi7QfPlfsPnMP2LjZ/RcwLeazRd+idUZbPILdC7gnMeZF2Ruu53+BiyeyjhMnW3TJSTKSLrPtgs0/QCpx5WxQ8CFsDX00VlFpkH3Z2S+ozCRYu8y2fwSuRVzn2aIDM4TkiR5cAGNps6zrSPdxXnJvM5pB88XbizZdBL0qVEoo8mVYr8WT98i01sOcu+9+D1McnFO04rBeYzWZKWY42dWRCKj54dSYhdav+prt0LYGmn2u3M72+EdjMemIE9LEQ+8UxBjZNUMHK9bzlY95+te1Gsh0g+yhFo0PU3nACW+w5/9NGv/LA0lHbko8bwhnJwS9mbi8xukY62al7DebYuxN5aufq/MVZUmNgvadsHgpQiSTdB5BVGIayqqxBkW6bDWmkwr6rzk4vQp9iYFKBkFLVthLgxeDHZaJ4GaRskIqLCGaWnZmxQYrVg0A1fPNngfhFGhJUtuUojJzqyUhZsxIj+uU+CmVkreHx8Yb6V5mWNTkRzHTZm9yYaYJOXdrbhf2hlw29hhRzvb4VHjsSnIjxriluY4XbacrjvOVr2Ykg+BdTtwtulYNkPiOCfv3CSOWG56YQm4C6PALYmdDYpA4VfYKqewBqPAhxq3gnj+kiztzBxXXiSoGt85XN+A22B8xOJQHgKJFmdrtDZbg5zciBT4+SemPH84JQJXz9acrHrONxLbLgIOYRSECLlRlLmlspqiMBRZRud8KsJx67pWFxllbpiVOXt1RpUWaTe9nS0K6H0QlWDK1BtPLtsFXC4qvLEI30pV80nROKagNDu3sx3exdgV5LeAVTtwuuo4WbWcrrrkDjZw3shSbtNJEKdCJScx8AGxcgwyK41BOlqjNRiF8h2WkBR1aa6pNOHlf0pnZjTzZ/H5Pp2a0dUf3Kr/iBL0aTRU7hx8j8cwmgpb71H9GbqekVvDQZ3zwuU5X/PMPkVu+PyVcz537YyzlaMZJLuu936baqzQFJki0wpjLdYolNJ0LtK5Aa0VVmumhaEqpXM9mJTM64y6yJgmQ6AiM6DEknTdDgxOmBKSsiFqvbqQpIx5nW+74nFs0DvPYtPL7PcetDOFJGHv3M52eDdhV5DfANbtwPGy5fp5w7WzDafrXtRxraQgj92vMXobFxSjUOx6hBtcWM2szOgGR68UzqutgEFffJp49C/xXjOkj8Zj2cSKgZzeZwyLNc6AsaUwJ5RCJ/qdQlzVgpdSXNNvF1UZjgv+hPd88MN88Jl9Ls5KXjpa85kXz3jlbM0mcXnbwW2VhRpZohVWJ99kvX0ua1WKNzIUmWZaSshnnWfsT3Iuzsutp0duRCSy6eSk0Luw7Ux1YrNMy5xL85K9VISt0Vva2emq29HOdvhNgV1BvgeEJ7vm6mnDKydrjlct6+Rz3A8el2KGbCrA0jUqVBS6lQgDNJmxGKNo+oGz1cDG9WgtxTnPNE3v6JyH8gninqI7f5nWa1pV4KJNFDSNIWB8jyWgqolE2SudQj0lsaI0GmUc3nssjkNOeJ6X+CBfZq+uaJ77d3n5ZMPP/to1jpcdTTew6T3tZoHv10Q/YIxFZ1N0Xoszm9ZM8owil4DPGKHKDPu1+P7WhWVelRxMc/YnOUVm0cl/YtU6ztcN3S0+zkqJSOTCtODivORgWlLYm254R6/jhrejne3wuGJXkLnp5na02PDK6YarpxuunG9YrAe6JEHunMTVRyT00mpFlczRx9Th0hpsCsxESae66T2bfmC9cjKiSA5om078J0ChtdCw1r2nczWh/gAxBsLqCEUgx6FxGAIah/WRmD2bbDWlQ5bUD4kx2r90iRdu/COe8y8xpQEiJ+YCn3nqj/PyL3yFszROaTvPEDzBdRi/RHlQGPAOwxmTAmb7lyjSwq3MNPMq53BWMk1c4b06Z38qydQmLeXW3cCqda+hkuVWc2FW8kRiYfQu0PSOG+fNjna2ww78JivIMUaawbNqJN34ZNly5WzN9fOGk1XPpnNbJVjwYRsnn2kxpKmKjFnitua5IdOJoaA0AZFfD26U1cqMs2kdm+S12w1idm5T7JBC0oo3yVM5xcbJ7s1oMtOR+4aY9HMpq4JocpTRkuoBFFZigZ65MOHJ/Zoqfy/xK4HNv/xxXu4UV+wLXNv7EIvTGb1b0A+RmJzftVIYL/4VBkdFy5wFM9+QrV6heOEPslfJCGJSZsJDNkpEGBPxnnAhsm5vinK2ghKEjjjyjEdP6GUzcLru7/gZ7dzOdvjNjMe2IPdOCu+yHVi3jnU7sGj7rRnP6brj+nlD7yKD9yIpTg5ldW4py4xZLTStzBisFoWYUpEQFT6Kj67zgd47hhTpPhoWrZoh+SvI0i3LpIvWWtM7x6aV2CThCsvjxhiZJGFFVBEXL+DW11BeYj8NHkzEzi9S15ZZVXB5v2R/UmCVaJc3neN42XAS38eVp/8dTpadpJi0nuAHMdHUwvEl+RPnfsUea+asqOiYsuaAMw78ksn7/8RWcFEXooarMvOaIjzOduXEIDPhOjdooxlcYNEMd/ycdrSzHXa4icemIB8vG46X3dZusellA7/pB5Yb8XZYtbKwCsmrt+9FdLBfZymIsmRWZxilKTJhPsQUJdT7QOdECDF4SRYenOdsNbDqBprBM/SezgsrwWh5jMJYjNE472mGwDD0hCiMgqo04gecCrE2isFHNoNwk02xT6ZAra6S+Zap8ew9+X4Onv+AKAxlUMu6GfAxslgPXD/fcLzqWHdDCgSV9ObgPcZaGUkoKK1mUubMq4xJd8yF4RUucMohZ5hkLmSrGSoTjnFdSD5dN3iuLZqUFShdsDGKTJEikWSsoLWS6KZbWBA7t7Mddrg3HpuCfOV0w29cOWfVDZytB1ZtL5zUIKo0iYjXVEVK/80MmZVOrkrLKoWS4jt4Np2nG3oGL8d7n0zGz9c9y1bsHQcX0UoM2IX6GpkUWRI5yOhhCIFuEEUcKLQxROcZvEics0xDUPQ+bmlrhbVYK8kX0/0n2Xvv89R5Rjj6HLz087Qv/zQmq4jPfiOr6fOJM9yz6TzO38zlC0EoZdZo0On1Z4ZJIeOGJ/YrLu/V7F/4IPYXfxbjJWEEQJmM/Ov/GNN5iQ+RReJKD0n6LGkfKok0XusTMdLOdm5nO+xw/3hsCvKXri/55S8fA5AZKbD7taHIhDObW8OkkCXU+DXpciOrtuds09M7seJUgAvSYZ+ues7WHd0geW+JLozViqyQOXCRG2ZFhjUKF0lzZC9ds4/0ThKb/cga0BJnpDQ4F1OAp6bODHkqYFXqIj3yeGevfA6ufBLtAxsOWA1TNl+5Ql9rej0hBJE/h5BCSgGbWYyWhOZJbpjXBZf3Kp48qLm8X1HlVpJADn8PMVMMv/x/ojbXmU1qsg//CVbP/l6+cmOFC2E7thDqntmKNEZ/kF3I5g47vHU8NgX54qzk0rxGa1nCjfPOMWjUGM3gxmj3fpsQ7bxn8CQRhGTCtYMnNbSEZHxe5BqNGNVPq4yLs5LCiHRjuXGsEmtBJLuBwTl6H3HOEUIkKum+NRqdEjwzY7by3zyTDr3MZObapRk4qRMdrn6Jzh/SkeMwyUzI4DfnxGkNJBc2rbfc3syI0u3SvOLpw4rnL864MC8gktzjZBQDUL/wrUy/6veybh2vdAPOR8yiJbfpppZc08YQz4dKO/vUj8LHPwbnL8Hes/Adfx6+9vse3OPvsMM7FI9NQX76cMK6c0wrOT4TYdn0LFvHyXotZjJRtvhjB/zi8Yq29/goqRcSMCodoErfl6UU5cNpwVOHNUWmeeV0wysnG1lmBUlXbn1g0w54LzS6YRDD9CCWEeioUnKIoco101IUaD5EYhTa2+ACZ2thP4QojmsugA+e6DQRs5VW+6Tii76HKNl8Y1BqOSzYP/8NnnZf5oPFOU+953vQX/X7aYfA6UrCVkOQxBOr5aRwtJCl21iAiwyMUtSl5aDO2ZsUbw/t7FM/Cv/3n4KhkT+fvyh/hl1R3uGxx2NTkHVSu50sO168scaNR3chExDTLFlrRZVJvPvnry6ZVFkK/ZTjeL1NkJAjfl0YnI8cLRp+9cVTFq1DdMqS97ZqB/ogHOUh5cUFB1GNAgi9dQ+rc0NVZKAiXe852/QMTpaGKkVkiMeFcJN9el02MwxG433PQIUjwyFFUQF6/Qr5ZJ/Z9BJPcY33v/x3eNa/yIw1m67m6Gf/Fn2jGZ78egYn7m7WGgqrcT6NS0p5vMyIE5sINoq3n3b28Y/dLMYjhka+vivIOzzmeGwK8rXzlpeOV9tMOWIErcityHrHjjSzwk4YvOeV0w3WaOZlvjW6MUahUbSDCBZO192WnaG1hHSuu4HBeVyIeBfoU7BpTCTim+Y6lirXW9ZB1wduLJvtLFmWY3KjsFa8hoWfLKMHosTTr5qBrnqevjkm+lEOAgbPjA0X/BHPrW/w7Ps/wuGv/W94f8KKKVd4KjnGZfArP0V1+euYVFYSQtL7NEqODyZShMesukeG85fe2NfvhN3IY4d3KR6bglzmUvQmpVDYRh5snqLdcyuUssFLTp738FVP7m2ToZ0Xv+DjZZ9CNj0uILNgNxqZyxjCe0mXGHzEx4CKIrAoclkWFlY6YqUiLsDpemDwwsgwiZVhkAI8rQyFNSQ3TKxWdIMUYXGH88QQCfmMEEB3Z0z9GYecJ5raOfssmfkV/rP/F19uSnregyJiGbAMTFmRdUfkkyK9H3orJpnXBZPSvnPEF3vPypjiTl+/H+xGHju8i/HYFORnL0xSt2e3S6fMKHonvsRn64FN77YJEb1z9C7SDI6zZXszIsloXOIvy5LP0w4DPkDXB3ovM2cCW3ZElma3mRHlHiia3uFSfpJRUGRa8uVyy16ZUZUWhcJ5CVftvWOxcayannU/4JKFbwyREAOltcwvXWSveoris3+PGUsuckxNh8dwg0voxlNUE/LmJKnuOmasKRgoJvuYw4lYWVbiHf2O9H/4jj9/e0EFyCr5+v1gN/LY4V2Mh1aQlVL/M/CvAddjjB96WM8zYl7lzKucEMWVbbkZWLbCzR2721EavemczH7dWITVNs14temki07Ku03Xs24Tt1fdpLypRCdTWpZfuZGg0BBjmldDkevtDHm/lp/PxYjzHq20RBU1A8tOeM29F2m1VgoVJcm5SmGfIriwXNwreLI8oWiviwMcOQU9FzhBVYfkv/27mP/CX2PizxjLbWFh/u3/NtMnZm9LPuBbwlg03+zI4UGMPHbY4RHhYXbIfxP4a8CPPMTn2GLTOc42Io0e5bzrdtgW3nU70CZbSZ8Wfj7KGCGkMcaycTTDQNOOHbQo8khFODHhsEqjtLAyCmvIMwNRGBUqQlGIGc68ypiUmcQ0uUAfJAlj3TrOm56u97gYGRLHWbh2EuZZ5KKOy61hVmY8c1hzcU9YHmH4Tqpf+psU/ixl6ilKA9Nv+kPY3/IRmEL+z/8HZsvPM53vk/+BP/fu6g6/9vve/M/7VkceO+zwCPHQCnKM8RNKqRce1uO/GjcWLS8erTjb9Kyani6JM8RrQlzWrBHF3qTIsUazbjtOVz0n64514+i8px8cIcp4QUzbxMFMGZEba8DalFydrCgDMXGfM+rckucaHRWOiA/SQbeD52TV0Z1dQ519mcF7OjMh1JehkFk2Gialpcozcq05mBW88MSc91yckFsrNwajGGa/j3WmGD79fzBdf5nJZI76pj9J9tXfxazKmH7z91J82x99u976dxbe6shjhx0eIR6bGfLxsuGzL58yuCBjAReICqxCimRmtnQ3HwJXjjd8+itnNMPA4KIURJX4vKn7VSkFw+iY0iwU07KgsJohdbOZ0RS5prCiepN/rzFG0zY9188bmnZAG0VcXKc9e4neF6KQ84GwuYEF5nsXJNqosLzviRm/470XeeZwwuBlBr7pxUC+G7w4qP3Wb6f4HX9AhCplxrzKKPPH5uN883irI48ddniEeORXsFLq+4HvB3j++eff9OOsmoFFM6AU5EYzqzKsTSbwVkYKy2bgC9fOOV51adHXiTjCKGwmhTTTitxailxRj0briNOZVlKInQ/kVuTZZWawWmEzg0HsNK+eNiyaAa3Fm6L3gU3jUWfHWB9ReDw5GQNTf8q0OebSV/8r/M73XuTr33cRF+Bk1XLltGHdDbS9RykJcr0wLci/+HGmP//fMlv+BtX8IuoPvIWC8zhSxN7KyGOHHR4hHnlBjjH+MPDDAB/+8Ifv5lH+utibiHOZ1QqbwkCdD5ytO87WPafrbksjk+Y2yqJMicFPmUnmW2lVklobQlr4OS9RQqhAmRlmk5LcQGYtMXGFT08bzpuewXu5AahIMwTpmoEy0/R+wJFTs+aQ6+yx4Bmu8HXuV3nuO/8My03PF6+vWLYDTSe+wrnVHEwLpqVlVubMvvBj1D/zp1AuHckXb4HWtaOI7bDDOwqPvCA/KDgfIURaF2jWHcvGse6HWzwrRPWWGc20thSZZd06JoWlLAzWaEorLm3rTlgZvROxR5lb9uqCqjAoxdYD4mzdpCLvIEmvFYreBYpMMytzurRQ1ErxhFmx56+zx5IXeImv49NMaLg2/Vp++YtHbG7xFa7TUnB/WshcuBT7Sz7xMXAPiNa1o4jtsMM7Cg+T9va3gW8HLiqlXgL+sxjj33hYz/fS8YqvnKzpkzHPGFtPipWfFhI/dHm/Yn8iJusvnqxJimU6Fzlve5rOY5Nybl7mkiGnQMXIppW06E03sOqcdNBaqG1aQ5lZJqW8petW+M7GaC7v19RFxsXZ+/nA536C3+Z/hZ6Cl3mKM3MJ/zX/FnQOrWBeZTwxr9hPXfFrLCsfJK3rQVPEHsfxxw47vI14mCyLP/6wHvtOcF460RCET2y0Zq+yVKVlv844nJbsTwusEh/kbpBYpGXbs2kdPga00lSZyKsnhaHKM0keaXvawYuEevBi4JMbylrYGtPSUhhLO3gWbc8wSCHen4gK7vLBhA89f8DTBx/k+BcCn/q5v0O/WUB9CL/te8nf9y1cnJVc3quYp8e8Kx4kretBPtZu/LHDDm8Zj83IosxNoqEpilzSnidFxoV5SV1klLlONpGaECKLZs0rp2sGF1LUvWFaZexVYtmJIo0lBq6eNagU7XRxXpIbQ2EVdZWjCZxvHMfrht4FjNLM6pz9uuDZixO++pl98sxw47zlF794AoffDN/9zQDsVTlPH9bymPfroPYgaV0P8rF2448ddnjLeGwKcmENdZklj17L4axgvy6YFGOEkiQcr1rHctPRDgFrDNMy42BSUmSaGMX7uHWe3kUgMCksz12YkmcixZ6XOXmuaPvA8aJlmbyDtZa/uzireM8TE56+MIUQuXq6ofc3d5W51Ty5X/H0wYRJmd39Bd3t+P8gaV0P8rF2CrkddnjLeGwK8qW9kovziguzgsNpgVGKwUVR7PWevhcT+W6Q9Aut4KmDSpI7lJKFmVKExEe+OJMlHsCmdUzrnBAix4uGK9cbNr3EMlktLIgn9kqePZxyMC3oXeDqyYYxTU4rOJgWPHVQ88S8en03tdc7/j9IWteDeqxHpZDbza13eIzw2BTkw1nFC5emoMSnOCBy5fONMC1cjNjkhxwlAG/LXOgGWQDOKsvThxPy1C1XeUY3iNz6laOVPFYvgaK5tRxOCy7NSy4fTJgWFh/gaNkBUoQnuZXIpP1KfJDvF+/G4/+jUMjt5tY7PGZ4bApymxKkfYysO4lpckECOTWgYsBFyX3Lrcb5SDt4pkXGey7NmFeZpIogDm6LpufLN5acrTsWm46mF1+LKjPM65yL84qL84JpmTM4z7J1KISuNi0znjyoOJgUb87M5914/H8UCrl3441rhx3ugcemIGdacb7pWHcO7wM+Qtd7fJT58qTK0SgZMxjNM4cVF2Y5LohVpjEikT5athwtWhatJDk3nRevY6vZm+RcmJVc3qspc0Pbi3NcmRkOK4k6OpyXzMrsrZm8v1sNct5uhdy78ca1ww73wGNTkF8+XXO0bCWbLiVHT0oRgBglcfUXpiWX90tAsWodTR+pco0BXjoShdxi03O66ekGj9WaujDsVQVP7FdcmJVYo8XgPkT2JgV1Ydmvc/Yn+YPzktgZ5Nwf3q03rh12uAsem4KslWbTeXKrOJwU1KWlsJa6NFyYlpSZpR2EZZFbQ10YTlYdLx4taXrH2bpn0fT0g3gVTwrD/qTkqYOK/arYRjPVueXizJKnxI291+MNvx7utZTaLavujd2Na4fHDI9NQb68X/K+yzOq3GCN2cqNIRnW+54qs6hcSSpHN3DtrOFk1XK66Qk+YhRMq4wLs4on9+tEmVNbW80smRXtT0TOfN9jibsV3ftZSo3/7uMfu/3rO+xuXDs8dnhsCnKVZ1yaV0zLnDLXNL1n3Q5k1lAV4ltxvGzRWrFqBq6er/ny9RU+RKzV7NU5l/cqLu6VTPKMshCO8ijYmJXZmxtL3Kvo3mspBTsGwf1g5+y2w2MENVLA3gn48Ic/HD/5yU++qX+72PS8dLxm0ztA3NXES7jHSwA1x8uWV042LDcdfZDs5mlleXKv4nBaMKsK5nVOkUkR1kq99bHEX/nQXeacz6Xl053ef3WP+ehz8NFfeXM/yw477PCocF/H6cemQ5ZQUXFZW7eO65ue/7+9e4+Rq6zDOP59ut3t7mzLbi90W9tqIRIuRihIEAIxCopAABNiCIgGI6SaYATvGBNEEv/ABKwkqGkULwmIiKiEEIRUjEYNWK6WS0NFkFpoy6XlUqjd7s8/zrtldjvd7kx79rw7+3ySycx5Z5p5cnL623Pe877vdE2fRgwFG155g41b32Tb9p0MpfHIs3u7GOjrZn5/jZkzOjmg1rWrC6Klbok9aVRUh9v7luz5ptRUHUHgiR42hbVNQR4cGmLrtu0MBUzvgMHBnTy98VW2vL69WFR+KOjsEPNq3SyYU2NBX41ad7EG8rCWuyXGog6InY3fO+RUeOSmxjelVl1V7giCHAufJ3rYFNc2BbmzYxpDEbz02pu8sOUttm3fweBQsGNnMGP6NBbPrTHQ38vc3hn0dHfSmbog9ku3xFj2VIwBnrobzrpuz4WxrBEEuRY+T/SwKa5tCvK657fy2H+28ObgIDsGhxjaOURvdxfvHuhhfl+N3u5OeuvWFx7ulpjZkxZ+L8ueuiWgKMJ7uilV5giCXAvfVO2mMUvapiC/uWOQrW9tp0Ni3qwu3jG7l1k9XdRmpF/bmFYU3ZmpW6Jnon4Q9JQr4LblNLx5t7fuh7JGEORa+DzRw6a4Eq7Rq/HOebN459yZHPWuuRy6aDYLZs9kfn9t14Lvc2bOYOn8WSycXZu4YgxFQT32M+x2k7XKCQx7KnBVF75Trij2Sz1P9LAppG0Kcn9tBkvmzWSgv8ZAX41ZPZ30dHYw0NfDQQOzmDure1e/8YQ781o4Z2XRfYGK57OuG//Z76O3FMPnruwvnh+9Zd/y5Fr4jjy32C+t7iezSa5txiG//PpbvJSWvpzwbokyjb4BB0XxHC5UrY6WyHGUhVn7mlrjkPtqXUTAAbWu6s6Ey1DWbD7PcDPLTttUro5p06rtlijLWDfg9laszWxSabPq1YbGugGX62gJM2uJC3LuxroBl+toCTNriQty7sYaeZDraAkza0nb3NRra1XM5psoHu1htosLclX2VyGazKMlcl1Tw6wi7rKownAh2vocEG8Xor1N+NjfE0Sq5lEiZiOUWpAlnSZpraR1ki4v87uaLlZVFrdmCtGunH3FmhjNFvG9qXI/eJSI2QilFWRJHcD1wOnAEcD5ko4o5cuaPeNs9aiCalIAAAZdSURBVAx1fxlvIRqRE3ZboGhfzyar3g8eJWI2QplnyMcB6yLi6Yj4H3Az8LFSvqnZS9+qL5XHW4ga5RxtX84mq94PHiViNkKZBXkRUL+W4vrUNoKk5ZJWS1q9efPm1r6p2Uvfqi+Vx1uIxpNnX84mq94PXkzIbIQyR1k0Wkxjt5WMImIlsBKKxYVa+qZm19Gtet3d8Q5X21POYft6Nln1foDJPUrEbD8r8wx5PbCkbnsxsKGUb2r20jeHS+Ujzy1+PfrKLcVzo6LUKOfw37n9cTaZw34ws13KPEP+B3CIpIOA/wLnAZ8o5ZuanSAxWSZUlJ1zsuwHsymi1PWQJZ0BrAA6gBsi4jtjfX5f1kM2M8tY9eshR8SdwJ1lfoeZWbvwTD0zs0y4IJuZZcIF2cwsEy7IZmaZcEE2M8uEC7KZWSZckM3MMuGCbGaWCRdkM7NMuCCbmWWi1LUsmiVpM/Bs1TnGaR7wYtUhmuC85ZtsmZ23XPV5X4yI0/b2D7IqyJOJpNURcWzVOcbLecs32TI7b7layesuCzOzTLggm5llwgW5dSurDtAk5y3fZMvsvOVqOq/7kM3MMuEzZDOzTLggm5llwgV5HCTdIGmTpDV1bXMk3SPpqfQ8u8qM9SQtkXSvpCckPSbp0tSeZWZJ3ZLul/RIyvvt1H6QpPtS3l9J6qo6az1JHZIeknRH2s42r6RnJP1T0sOSVqe2LI8HAEn9km6V9GQ6jk/IPO+had8OP16VdFmzmV2Qx+dnwOhB3ZcDqyLiEGBV2s7FIPDliDgcOB64RNIR5Jt5O3ByRBwFLANOk3Q8cDXwvZT3FeCiCjM2cinwRN127nk/FBHL6sbG5no8AHwfuCsiDgOOotjP2eaNiLVp3y4D3gdsA35Ls5kjwo9xPIClwJq67bXAwvR6IbC26oxjZP898JHJkBmoAQ8C76eY5TQ9tZ8A/KHqfHU5F6f/YCcDd1D8qnDOeZ8B5o1qy/J4AA4A/k0adJB73gb5TwX+2kpmnyG3biAingdIz/MrztOQpKXA0cB9ZJw5Xf4/DGwC7gH+BWyJiMH0kfXAoqryNbAC+BowlLbnknfeAO6W9ICk5akt1+PhYGAz8NPUJfRjSb3km3e084BfptdNZXZBbmOSZgK/AS6LiFerzjOWiNgZxeXeYuA44PBGH5vYVI1JOhPYFBEP1Dc3+GgWeZMTI+IY4HSKLqwPVB1oDNOBY4AfRsTRwBtk1D0xlnTf4Gzg1638exfk1m2UtBAgPW+qOM8IkjopivGNEXFbas46M0BEbAH+RNH33S9penprMbChqlyjnAicLekZ4GaKbosV5JuXiNiQnjdR9G0eR77Hw3pgfUTcl7ZvpSjQueatdzrwYERsTNtNZXZBbt3twIXp9YUU/bRZkCTgJ8ATEXFt3VtZZpZ0oKT+9LoH+DDFTZx7gY+nj2WTNyK+ERGLI2IpxeXpHyPiAjLNK6lX0qzh1xR9nGvI9HiIiBeA5yQdmppOAR4n07yjnM/b3RXQbOaqO8AnwyPt4OeBHRR/vS+i6DNcBTyVnudUnbMu70kUl8uPAg+nxxm5ZgaOBB5KedcAV6T2g4H7gXUUl4Azqs7aIPsHgTtyzptyPZIejwHfTO1ZHg8p2zJgdTomfgfMzjlvylwDXgL66tqayuyp02ZmmXCXhZlZJlyQzcwy4YJsZpYJF2Qzs0y4IJuZZcIF2dqKpJB0Td32VyRdWbe9PK0g9mRaYe6kuvdulLRW0pq0wl/nBMe3Kc4F2drNduAcSfNGv5GmPH8WOCmKVcQ+B9wkaUH6yI3AYcB7gR7g4omJbFZwQbZ2M0jxW2ZfbPDe14GvRsSLABHxIPBz4JK0fWckFBM8Fk9MZLOCC7K1o+uBCyT1jWp/D/DAqLbVqX2X1FXxKeCu0hKaNeCCbG0nipXtfgF8YRwfF7uvyvYD4M8R8Zf9nc1sLC7I1q5WUKw50lvX9jjFrznUOya1AyDpW8CBwJfKDmg2mguytaWIeBm4hZE/o/Rd4GpJcwEkLQM+TXFGjKSLgY8C50fEEGYTbPreP2I2aV0DfH54IyJul7QI+JukAF4DPhnpFx2AHwHPAn8vVjDltoi4aoIz2xTm1d7MzDLhLgszs0y4IJuZZcIF2cwsEy7IZmaZcEE2M8uEC7KZWSZckM3MMvF/4w51ZNU73foAAAAASUVORK5CYII=\n",
"text/plain": [
"
"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"sns.lmplot('NO2', 'SO2', data = no2_so2_boot,\n",
" # Tell seaborn to a regression line for each sample\n",
" hue = 'sample', \n",
" # Make lines blue and transparent\n",
" line_kws = {'color': 'steelblue', 'alpha': 0.2},\n",
" # Disable built-in confidence intervals\n",
" ci = None, legend = False, scatter = False);\n",
"\n",
"# Draw scatter of all points\n",
"plt.scatter('NO2', 'SO2', data = no2_so2);"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The outliers appear to drag down the regression lines as evidenced by the cluster of lines with more severe slopes than average. In a single plot, you have not only gotten a good idea of the variability of your correlation estimate but also the potential effects of outliers."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Lots of bootstraps with beeswarms\n",
"As a current resident of Cincinnati, you're curious to see how the average NO2 values compare to Des Moines, Indianapolis, and Houston: a few other cities you've lived in.\n",
"\n",
"To look at this, you decide to use bootstrap estimation to look at the mean NO2 values for each city. Because the comparisons are of primary interest, you will use a swarm plot to compare the estimates."
]
},
{
"cell_type": "code",
"execution_count": 26,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
city
\n",
"
year
\n",
"
month
\n",
"
day
\n",
"
CO
\n",
"
NO2
\n",
"
O3
\n",
"
SO2
\n",
"
\n",
" \n",
" \n",
"
\n",
"
136
\n",
"
Cincinnati
\n",
"
2013
\n",
"
5
\n",
"
121
\n",
"
0.640
\n",
"
33.0
\n",
"
0.055
\n",
"
1.85
\n",
"
\n",
"
\n",
"
137
\n",
"
Cincinnati
\n",
"
2013
\n",
"
5
\n",
"
122
\n",
"
0.315
\n",
"
23.0
\n",
"
0.043
\n",
"
4.10
\n",
"
\n",
"
\n",
"
138
\n",
"
Cincinnati
\n",
"
2013
\n",
"
5
\n",
"
123
\n",
"
0.260
\n",
"
13.0
\n",
"
0.039
\n",
"
4.75
\n",
"
\n",
"
\n",
"
139
\n",
"
Cincinnati
\n",
"
2013
\n",
"
5
\n",
"
124
\n",
"
0.245
\n",
"
17.0
\n",
"
0.051
\n",
"
4.45
\n",
"
\n",
"
\n",
"
140
\n",
"
Cincinnati
\n",
"
2013
\n",
"
5
\n",
"
125
\n",
"
0.230
\n",
"
12.0
\n",
"
0.032
\n",
"
0.85
\n",
"
\n",
"
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
\n",
"
\n",
"
8673
\n",
"
Vandenberg Air Force Base
\n",
"
2015
\n",
"
5
\n",
"
137
\n",
"
0.000
\n",
"
1.0
\n",
"
0.044
\n",
"
0.50
\n",
"
\n",
"
\n",
"
8674
\n",
"
Vandenberg Air Force Base
\n",
"
2015
\n",
"
5
\n",
"
138
\n",
"
0.000
\n",
"
2.0
\n",
"
0.043
\n",
"
1.00
\n",
"
\n",
"
\n",
"
8675
\n",
"
Vandenberg Air Force Base
\n",
"
2015
\n",
"
5
\n",
"
139
\n",
"
0.000
\n",
"
0.0
\n",
"
0.041
\n",
"
0.50
\n",
"
\n",
"
\n",
"
8676
\n",
"
Vandenberg Air Force Base
\n",
"
2015
\n",
"
5
\n",
"
140
\n",
"
0.000
\n",
"
0.0
\n",
"
0.044
\n",
"
0.50
\n",
"
\n",
"
\n",
"
8677
\n",
"
Vandenberg Air Force Base
\n",
"
2015
\n",
"
5
\n",
"
141
\n",
"
0.000
\n",
"
0.0
\n",
"
0.041
\n",
"
0.50
\n",
"
\n",
" \n",
"
\n",
"
736 rows × 8 columns
\n",
"
"
],
"text/plain": [
" city year month day CO NO2 O3 SO2\n",
"136 Cincinnati 2013 5 121 0.640 33.0 0.055 1.85\n",
"137 Cincinnati 2013 5 122 0.315 23.0 0.043 4.10\n",
"138 Cincinnati 2013 5 123 0.260 13.0 0.039 4.75\n",
"139 Cincinnati 2013 5 124 0.245 17.0 0.051 4.45\n",
"140 Cincinnati 2013 5 125 0.230 12.0 0.032 0.85\n",
"... ... ... ... ... ... ... ... ...\n",
"8673 Vandenberg Air Force Base 2015 5 137 0.000 1.0 0.044 0.50\n",
"8674 Vandenberg Air Force Base 2015 5 138 0.000 2.0 0.043 1.00\n",
"8675 Vandenberg Air Force Base 2015 5 139 0.000 0.0 0.041 0.50\n",
"8676 Vandenberg Air Force Base 2015 5 140 0.000 0.0 0.044 0.50\n",
"8677 Vandenberg Air Force Base 2015 5 141 0.000 0.0 0.041 0.50\n",
"\n",
"[736 rows x 8 columns]"
]
},
"execution_count": 26,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"pollution_may = pollution.query(\"month == 5\")\n",
"pollution_may"
]
},
{
"cell_type": "code",
"execution_count": 27,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"