{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"![title](header.png)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# What is Statistical Analysis?"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Statistical Analysis in Python lets us draw formal conclusions from data using mathematical tests. These tests depend on a series of assumptions to work, but if they are met, can give us some powerful insight."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Why Python for Statistics?"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"R - another programming language - is purpose built, and made for statistics. So why use Python instead of R? Python allows us more flexibility with our data; Python has more options for importing data, working with controllers and image analysis that makes it more versitile than R. Prehaps more importantly is the size of the community - Python has a very large following and sees regular updates and improvements to it's ecosystem, moreso than R."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# What libraries?"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"There are a number of libraries that are useful for statistical analysis in Python:\n",
"\n",
"* numpy and pandas give us a nice format for us to analyse our data in.\n",
"* matplotlib and seaborn give us cool visualisation options.\n",
"* scipy and statsmodels give us good frequentist options for analysis.\n",
"* PyMC gives us good bayesian tools for analysis."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Frequentist vs Bayesian statistics"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"These are two different philosophies about what statistics and probability mean, resulting in different methods and outcomes. A frequentist believes in probability being the result of \"long term\" distributions, and a bayesian believes in using new data to reach new conclusions. A good analogy for this is what would happen if a frequentist and a bayesian lost their phone in their house, and rung it to find it - a frequentist would trust the sound first, whereas a bayesian would trust their knowledge of where they normally leave their phone.\n",
"\n",
"I recommend trying both out and seeing which one works for you - Python has more support for frequentist analysis but that shouldn't put you off trying bayesian analysis!"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# What Problems Is This Useful For?"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Strict statistical testing is useful for academics to test hypothesis in experiments and other studies - if we want to definitively test if two means are equal (eg does an antidepressant improve happiness scores over a control group?), how correlated two variables are (eg the concentration of a substance versus it's fluorescence), and if a group of means are different (eg different material elasticities)."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.6.0"
}
},
"nbformat": 4,
"nbformat_minor": 2
}