Lecture 9: Intro to Python

What is R?

R is a programming language specifically designed for statistics and data analysis
- Objects for storing data, and functions for interacting with data, are fundamental
- R is very good at graphics and visualization
- R is easily extended. Users can write and share their own functions and packages
We can interact with R through IDEs like RStudio

Who uses R?

R is widely used in statistical research and academia
R is also widely used in applications of statistics, such as
- Biology and bioinformatics
- Ecology, forestry, and environmental science
- Psychology
- Sociology
R is used in a variety of areas outside research, by government agencies, pharmaceutical companies, insurance companies, etc.

What other options exist?

SAS
Stata
SPSS
Excel
Python
Julia
Matlab
Many others…

What is Python

Python is a general-purpose programming language
Like R, python has a wide range of packages to extend functionality
Certain Python packages allow for sophisticated data analysis and modeling
- SciPy, NumPy
- scikit-learn, statsmodels, pytorch
- pandas
- matplotlib

R vs. Python

My own, personal, preferences:

R is good for

Data visualization and wrangling
Classical statistics
Statistical inference
New statistical methods

Python is good for

General-purpose programming
Challenging data types (e.g. images)
Prediction and machine learning

Why do we teach R?

Excellent support for the material covered in a statistics degree
Benefits to using a consistent language across courses
Valuable for a wide variety of future careers
The primary research tool for most (if not all) the faculty

A taste of Python

Recall our code from the first class:

M <- 10 # number of people at the party
hats <- 1:M # numbered hats
nsim <- 10000 # number of simulations
results <- rep(0, nsim) # vector to store the results

for(i in 1:nsim){
  randomized_hats <- sample(hats, M, replace = FALSE)
  results[i] <- sum(randomized_hats == hats) > 0
}

mean(results)

[1] 0.6296

A taste of Python

Here is the same code, written in Python

import numpy as np

M = 10 # number of people at the party
hats = np.arange(M) # numbered hats
nsim = 10000 # number of simulations
results = np.zeros(nsim) # to store the results

for i in range(nsim):
    randomized_hats = np.random.choice(hats, M, replace = False)
    results[i] = np.sum(randomized_hats == hats) > 0

np.mean(results)

0.634

What similarities and differences do you notice?

Step 1: representing the hats

import numpy as np

M = 10 # number of people at the party
hats = np.arange(M) # numbered hats

hats

array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

hats[0]

hats[1]

hats is a 1-dimensional array (similar to a vector in R)
Python is 0-indexed: the first entry is hats[0]

Step 2: everyone draws a random hat

import numpy as np

M = 10 # number of people at the party
hats = np.arange(M) # numbered hats

randomized_hats = np.random.choice(hats, M, replace = False)

randomized_hats

array([6, 8, 5, 2, 1, 0, 4, 7, 9, 3])

np.random.choice works like R’s sample function
Booleans in Python are True and False (as opposed to TRUE and FALSE, or T and F)

Step 3: check who got their original hat

import numpy as np

M = 10 # number of people at the party
hats = np.arange(M) # numbered hats

randomized_hats = np.random.choice(hats, M, replace = False)
randomized_hats

array([5, 6, 0, 3, 1, 8, 9, 7, 2, 4])

randomized_hats == hats

array([False, False, False,  True, False, False, False,  True, False,
       False])

np.sum(randomized_hats == hats)

NumPy arrays allow for “vectorized” operations, like in R

Step 4: iteration

import numpy as np

M = 10 # number of people at the party
hats = np.arange(M) # numbered hats
nsim = 10000 # number of simulations
results = np.zeros(nsim) # to store the results

for i in range(nsim):
    randomized_hats = np.random.choice(hats, M, replace = False)
    results[i] = np.sum(randomized_hats == hats) > 0

np.mean(results)

0.6282

range(nsim) is similar to 1:nsim in R
We don’t use the curly braces { }. Instead we use whitespace (four spaces is standard, you just have to be consistent)

Using Python through RStudio

You can make Python chunks in Quarto documents, just like R chunks:

```{python}

```

Class activity

https://sta279-f23.github.io/class_activities/ca_lecture_9.html