nakatsuma/probability_and_statistics

Course materials for PROBABILITY AND STATISTICS A/B

PROBABILITY AND STATISTICS A/B (MATHEMATICAL STATISTICS)

Course materials for PROBABILITY AND STATISTICS A/B (MATHEMATICAL STATISTICS)

Teruo Nakatsuma (Faculty of Economics, Keio University, Japan)

---

• How to set up Python and necessary packages
  • Step 1: Installing Anaconda
  • Step 2: Creating an environment
  • Troubleshooting about installation
    • 1. Retry installation
    • 2. (macOS) Installing Command Line Tools for Xcode
• How to start Jupyter Notebook
  • Troubleshooting about Jupyter Notebook
• Jupyter Notebooks and related files in notebook-a
• Jupyter Notebooks and related files in notebook-b

---

How to set up Python and necessary packages

I strongly recommend using Anaconda. It can install Python along with numerous essential packages at once and allows us to manage those packages flexibly.

Step 0: For Windows PC users only

Download and install Microsoft Visual C++ Redistributable. There are three versions:

• ARM64
• X86
• X64

If your PC's CPU is manufactured by Intel or AMD, the X64 version is suitable.

Step 1: Installing Anaconda

1. If you have an older Anaconda on your PC, uninstall it completely by folloiwng instructions.

2. Download an Anaconda installer (Windows, macOS or Linux) from here. Choose an installer for your OS. For macOS, you should check whether Intel chip or Apple Silicon chip is used and download the corresponding installer.

3. Doubleclick the installer and follow the instructions on the screen. Do not change the default settings.

Step 2: Creating an environment

Start Anaconda Powershell Prompt (Windows) or Terminal (macOS, Linux) and type

conda update conda

This will update conda (package manager) in Anaconda.

(Optional) To update all packages, type

conda update -n base --all

Then type

(Windows)

conda create -n bayes -c conda-forge jupyterlab seaborn bokeh jupyter_bokeh pymc python-graphviz

(macOS)

conda create -n bayes -c conda-forge jupyterlab seaborn bokeh jupyter_bokeh pymc numpyro python-graphviz

This will create the environment for PyMC. Then type

conda activate bayes

(Windows) Install numpyro by typing

pip install numpyro

Finally, type

python -m ipykernel install --user --name bayes --display-name "Python (Bayes)"

Now you are ready for Python!

---

Troubleshooting about installation

1. Retry installation

If you encounter any errors during the installation process, go back to the default environment by typing

conda deactivate

and remove bayes by typing

conda env remove -n bayes

Then redo Step 2.

2. (macOS) Installing Command Line Tools for Xcode

In case the computer says Command Line Tools for Xcode is missing, install it as follows.

1. Install Xcode from App Store.

2. Start Xcode. If a pop-up window asks you to install additional tools, follow the instruction. Quit Xcode.

3. Start Terminal and install Command Line Tools for Xcode by typing

sudo xcode-select --install

If asked, type your login password.

---

How to start JupyterLab

Method 1: From the command line

Start Anaconda Powershell Prompt (Windows) or Terminal (macOS, Linux) and type

jupyter lab --port=8888

Your default browser will pop up.

Method 2: From Anaconda Navigator

Start Anaconda Navigator. You may find it in Start Menu (Windows) or Launchpad (macOS). Then click the icon named JupyterLab.

Troubleshooting about JupyterLab

1. For a Bokeh interactive plot to work properly, the Jupyter Notebook server must use port 8888 which is set by default. In case this port is occupied by another Jupyter Notebook server, you need to stop it by typing

jupyter server stop

before you open a new JupyterLab session. If this does not work, reboot your PC.

2. PyMC may fail when you start your JupyterLab session in the bayes environment. If so, you must start it in the base environment. If the prompt starts with (bayes), type

conda deactivate

Then it will switch back to the base environment.

Jupyter Notebooks and related files in notebook-a

file name	description
Cholera.csv	London cholera pandemic data
Mroz.csv	US women's labor participation data
StrikeDur.csv	strikes duration data
USStocksSW.csv	monthly US stock returns data
ar1_process.ipynb	convergence of the AR(1) process
cholera.ipynb	Bernoulli model of the cholera data
example_bernoulli.ipynb	posterior inference on Bernoulli dist.
example_exponential.ipynb	posterior inference on exponential dist.
example_normal.ipynb	posterior inference on normal dist.
example_poisson.ipynb	posterior inference on Poisson dist.
exponential_duration.ipynb	exponential duration model
Housing.csv	sales prices of houses
housing_price.ipynb	hedonic price model of houses
labor_participation.ipynb	logit model of labor participation
logit.ipynb	PyMC example of logit model
poisson_regression.ipynb	PyMC example of Poisson regression model
probit.ipynb	PyMC example of probit model
prussian.csv	Prussian army horse kick data
regression.ipynb	PyMC example of regression analysis
ships_damage.ipynb	Poisson regression model of ships damage
ships.csv	ships damage data
strikes_duration.ipynb.	strikes duration
wage_education.ipynb	relationship between wage and education

---

Jupyter Notebooks and related files in notebook-b

file name	description
bivariate_distribution.ipynb	examples of bivariate distributions
large_sample.ipynb	consistency and asymptotic normality
markovchain.ipynb	Markov chain
probability_distribution.ipynb	examples of probability distributions
python_introduction.ipynb	simple example of Bayes' theorem
skewness_kurtosis.ipynb	skewness and kurtosis

---