Exploring Probability Distributions in Python with SciPy
Exploring Probability Distributions in Python with SciPy
Whether you are modeling test scores, financial risk, or daily website traffic, understanding probability distributions is a foundational skill in data science and statistics. Python’s scientific stack makes working with these distributions intuitive and visual.
Specifically, the
scipy.stats library provides a comprehensive suite of pre-built continuous and discrete distributions—including Normal, Binomial, Poisson, and Uniform distributions.Here is a hands-on walkthrough of how to define, visualize, and calculate key probabilities using the Normal Distribution in Python.
1. Setting Up the Distribution
A normal distribution (or bell curve) is defined by two parameters:
- Mean ($\mu$ or
loc): The central tendency or peak of the curve. - Standard Deviation ($\sigma$ or
scale): The spread or dispersion of the data.
Using
scipy.stats.norm, you can instantiate a frozen distribution with your chosen parameters:Python
import numpy as np
import matplotlib.pyplot as plt
from scipy.stats import norm
# Define a normal distribution with mean = 75 and standard deviation = 8
distribution = norm(loc=75, scale=8)
2. Visualizing the Probability Density Function (PDF)
The Probability Density Function (PDF) represents the relative likelihood of a continuous random variable taking on a given value.
To plot it:
- Generate an array of evenly spaced numbers covering roughly $\pm 3.5$ standard deviations from the mean (e.g., between 45 and 105).
- Calculate the corresponding density values with
.pdf(x). - Use Matplotlib to plot the curve and shade regions of interest using
plt.fill_between().
Python
# Generate x-values spanning the distribution
x = np.linspace(45, 105, 300)
y = distribution.pdf(x)
# Plot the distribution
plt.figure(figsize=(9, 4.5))
plt.plot(x, y, linewidth=3)
# Highlight values at or below the mean (x <= 75)
plt.fill_between(x, y, where=(x <= 75), alpha=0.25)
plt.title("Normal Distribution: ฮผ = 75, ฯ = 8", fontsize=15, weight="bold")
plt.xlabel("Value")
plt.ylabel("Density")
plt.tight_layout()
plt.show()
The resulting plot gives you a clear visual of the classic symmetrical bell curve centered at 75, with the lower half neatly shaded to represent 50% of the total area under the curve.
3. Calculating Probabilities: CDF and PPF
Beyond visualization,
scipy.stats gives you straightforward methods to solve practical statistical questions.Cumulative Distribution Function (CDF)
The CDF answers: What is the probability that the random variable $X$ is less than or equal to a specific value $x$?
- $P(X \le 70)$:Python
prob_less_70 = round(distribution.cdf(70), 4) print("P(X ≤ 70):", prob_less_70) # Output: 0.2660This indicates roughly a 26.6% chance of observing a score of 70 or lower. - $P(X > 70)$:Because the total probability under the curve equals 1, the probability of exceeding 70 is simply the complement:Python
prob_greater_70 = round(1 - distribution.cdf(70), 4) print("P(X > 70):", prob_greater_70) # Output: 0.7340
Percent Point Function (PPF)
The PPF (the inverse of the CDF or quantile function) answers the reverse question: What value corresponds to a specific cumulative probability or percentile?
- Finding the 90th percentile:Python
percentile_90 = round(distribution.ppf(0.90), 2) print("90th percentile:", percentile_90) # Output: 85.25A value of approximately 85.25 cuts off the top 10% of the distribution from the bottom 90%.
Summary Cheat Sheet
| SciPy Method | Purpose | Typical Question Answered |
norm(loc, scale) | Initializes the distribution | "Set up a normal curve with mean 75 and standard deviation 8." |
.pdf(x) | Probability Density Function | "What is the relative likelihood height at value $x$?" |
.cdf(x) | Cumulative Distribution Function | "What is the probability that $X \le x$?" |
.ppf(q) | Percent Point Function (Inverse CDF) | "What value marks the $q$-th percentile?" |
By swapping out
norm for binom, poisson, or uniform, you can apply these same core methods across virtually any common statistical distribution in Python.
Comments
Post a Comment